AMD ZEN
- Home
- Board: /g/ - Technology
- Reading: AMD ZEN
>Zen's main focus will be on increasing per-core performance rather than core count or multi-threading performance
>DDR4 support and a 95W TDP. Zen does not support DDR3
>As of November 2015, AMD reported that its Zen microprocessors had been tested and "met all expectations" with "no significant bottlenecks found".
>expected for release in October 2016
Will it bomb?
>>51433589
>October 2016
nigga by that time people will be running kabylake with 6 core i5s and 8 core i7s
>>51433589
>8core Haswell performance with no 3rd-world tier iGPU bearing on the TDP, all for ~$300
>Implying you wouldn't buy this
How's it compare with Nvidia's 3D super-CPU?
>>51433620
That's what I'm thinking too. If it had launched this year it MIGHT have had some relevance
>>51433620
kaby lake wont bring a performance increase.
>>51433621
By then you'll be able to buy a used 5960X for $300
>>51433589
>October 2016
Yeah, no.
Think more middle of the year, they're already making them.
>>51433651
Why do people mske post like these
>>51433683
i dont know, just look at performance increase over the last 4 years from Intel.
How many cores will it have? 8 wont be enough to justify much more than $200-$300. If they have a 16 core opteron for $600 I might be interested but most likely there will be no reason to use zen instead of the 10 core i7.
>>51433589
Zen is supposed to have hyperthreading, right?
>40% more IPC
That's about the added throughput hyperthreading gives you in a near-perfect (i.e. marketing) scenario.
Hmmm....
>"It is the first time in a very long time that we engineers have been given the total freedom to build a processor from scratch and do the best we can do," Plummer said. "It is a multi-year project with a really large team. It’s like a marathon effort with some sprints in the middle. The team is working very hard, but they can see the finish line. I guarantee that it will deliver a huge improvement in performance and (low) power consumption over the previous generation."
cautiously optimistic
2016 can't come soon enough
arr rook same
>>51433589
>>51433621
>>51433668
>>51433745
I'm definitely waiting for Zen, but I'm using an fx 8350 on a crosshair IV mobo. I can barely overclock it from all my tries and it gives me lots of weird issues with ram above 1333mhz. I think because it's a 2009 mobo that barely supports bulldozer let alone pilediver.
Hold me tight MADbros, it's gonna be a long year...
>>51433711
Why? What do you need all the cores for?
>>51433683
Intel won't be making the switch to a smaller node until late 2017 at the earliest, so we can't expect much of an improvement until then.
>>51433711
Looking at core count in isolation is plain retarded, senpai.
>>51433798
>tfw still phenom ii 955
it's still running strong on 4ghz, but i'm yearning for an upgrade. it's been like 5 years
>>51433661
>buying used components
>>51433651
yfw 'tic toc' is just intel pestering amd
>tick tock, amd.
>tick
>tock
I'm hoping Zen would be good to give Intel some competition. But if it's shit I'll just buy an Intel cpu. No problemos.
By the time Zen comes out, I'll probably be in the market for a new CPU. Hoping it's good because I fucking hate intel.
>>51433589
>October 2016
jesus, the last relevant cpu they released was 8320... in October of FUCKING 2012
effectively 4 years between launches, if it doesn't give intel a run for it's money, amd is very much done
>>51433850
>tfw Phenom II 940
>on asus crosshair
>A-M-fucking-2-period from 2007
>>51433809
>Why? What do you need all the cores for?
3d rendering. Current xeons are no good due to low clockspeeds. I need the highest single threaded performance and highest mt performance in the same chip. Xeons are ok for render farms but on workstations you still encounter single threaded tasks in 3d programs.
The fastest i7 is actually more productive than their $4000 18 core xeon in this situation. If they would just sell unlocked xeons it wouldn't be an issue as I could have a 300 watt 4.4ghz 12 core house fire.
>>51434049
>Prefering to support intel unironically
SJW's pls go
>ZERO BOTTLENECKS
INTEL ON SUICIDE WATCH
>>51434061
I have one on a Nforce 570 SLI from 2006.
Intelfags are going to do the biggest switchover in history. Literally /g/ is going to swear on their cpu's grave that they always backed AMD.
>>51434121
hey there anita, time to get to work on your next video
>>51434199
lol keep dreaming. i have no bias against amd cpus because cpus either work or dont and you dont have to rely on horrendous drivers.
I'll switch over to amd if it's good but by q3 2016 people will be running 10 core boardwell-e chips and 6 kaby lakes.
Last time I checked zen barely gets haswell performance.
>>51434069
Anon, have you ever heard of SIMD?
>>51434284
This isn't reddit. Stop with your shitposting
>>51433589
>Zen does not support DDR3
>>51434049
>>51434121
>>51434284
>implying these people are white and not muslim
>>51434330
>unironically buying obsolete technology
>>51434358
>>51434358
>dat butthurt
Go back to your safe space nigger
>>51434330
Neither does skylake
>>51434358
You dumb nigger.
>>51434399
>>51434413
>being inbred white trash unironically
>>51434086
I'm afraid they're just saying that so that they can rush the release while building up hype
AMD are masters of hype and under-delivering
>>51433589
pls be good
i'm holding off my upgrades for this
yet i'm not optimistic at all
>>51433589
so... less coars?
>>51433620
why buy kaby when you haven't bought skylake? It's pretty much the same.
>>51433763
>I guarantee that it will deliver a huge improvement in performance and (low) power consumption over the previous generation."
As if that's difficult.
>>51434330
ddr3 will be dead tech in q4 2016.
heck, they're using lpddr4 in smartphones already.
>>51433589
Will this finally be a worthy upgrade from my 965?
>>51433621
There a pretty big risk they will only make apus. Plus they will be overpriced.
>>51433620
Except AMD chips will still give you 80% of the performance for 50% of the cost
this zen really reminds me of the fury hype
>the fury will beat the 980
>it will be on par with titan x
look where we are now
>>51434069
There are dual socket lga2011 mobos that allow you to OC xeons.
>>51434330
Pretty sure they said at one point it would support both
>>51434830
>lpddr4
>as slow as DDR2
Phones are truly the pinnacle of innovation.
>>51434924
We are now at the point where Fury is better than Titan X.
Did Intel's plan intend for about four tocks in a row? Is there any point in upgrading from a 3570k?
>>51433589
>Zen does not support DDR3
Dissapointing.
>BUT IT WON'T BE AS GOOD AS INTEL
Nigger who cares? I'll buy Zen as long as AMD can finally provide i5 performance so I can both not support ze jews AND experience great frames in my jewish media.
>>51434413
>>51434399
>>51434347
>>51434326
>>51434217
What did the deleted posts say?
Why can't the mods just tag them as faggots or osmething without deleting
>>51434924
The fury does beat the 980, the fucking 290x beats that piece of shit
>>51435087
>>being inbred white trash unironically
>>51433589
>the left axis has no units whatsoever
That triggers the fuck out of me.
>>51434264
>Last time I checked zen barely gets haswell performance
>I
I wasn't aware AMD Engineers were so skeptical of their own products. How is AMD doing? Since you clearly work there.
>>51433798
We'll get there Anon.
Just don't fall too hard if it tanks. We'll all have to endure another 10 years of Intel slavery before ARM gets any good.
>2015
>there are unironically people who still use AMD
hahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahahaha
>>51435342
I purchased my 8350 in 2012.
Wouldn't buy AMD right now, but no ragrets. I'll wait for Zen.
>>51434480
>unironically being a niggerlover
>unironically deleting your own posts when you got BTFO
>>51435377
>being inbred white trash unironically
>>51435366
I bought mine in 2013 I think. I'm glad I bought it instead of Intel. It still is serving me better than I imagine. I can run anything I want without a hiccup and I paid half of what I would have for Intel and still don't even need to upgrade. Definitely going AMD again.
>>51434941
They're old ones though aren't they? The most they can do is 2x 6 core, they're nearly impossible to find, cost $1000 per board and struggle to beat single newer 12 core xeons and 8 core i7s.
>>51435873
Asus made one for 2011v3. Gold in colour if I remember correct. Dunno what CPUs it supports though. Cost ~$600 or something.
>>51435342
> he hasn't been using AMD CPUs since the K5
>>51433620
>Kabylake
>6 core i5s
kek
It's pretty much confirmed Kabylake will be quadcore only. Intel will not have a mainstream 6 core processors in the foreseeable future.
So how big is this thing and why is the desktop chip only 8 cores?
>>51436395
That looks like a monstrous apu.
>>51435273
the main thing wrong with it is the typical "makes it look bigger than it is" tactic
the points suggest a 3-fold increase at a flance, but the "40%" shows that the graph doesn't begin at the assumed "0"
>>51436649
it's an SoC, it contains sata and ethernet (GbE)
>>51435960
These boards can only OC the BCLK, which you can get to around 105Mhz.
18 cores running at 2.7Ghz is still pretty good for what it is.
>>51433620
And AMD will most likely be insolvent
>Book Value Per Share of AMD: ($0.43)
>>51433589
>was thinking of upgrading
>was ashamed that I was going to buy Intel
>I thought I was forced to cück AMD
>tfw AMD finally has a promising CPU
>>51434199
This isn't AMD overtaking the Intel performance crown for less money like 1999-2002. This is AMD catching up to what will be 2 year old INTC performance.
I'm pulling for them, I really am. The only reason INTC started making real progress with their chips again is due to AMD beating them up for less money, but AMD is insolvent and these chips will only sell at a deep discount to the INTC chips. Doesn't look good for them.
They should split off the graphics business (buying ATI is what tanked the company; took on too much debt) and hopefully another company will buy AMD and infuse a proper R&D budget. Rumors had MSFT looking into it but said that would cause too many problems with their close partner INTC. I think MSFT should go for it now that they're in the hardware manufacturing game with the surfaces and gaming consoles.
>>51437022
Somewhat similar to me:
>Buying a new rig to do GPGPU work on the desktop rather than over SSH on a cluster machine after having nothing but Macbook Pro's since 2007
>AMD hasn't been competitive in the higher end desktop market for a while after their strategy of focusing on high theoretical performance trough parallelism failed due to software not keeping up with their hardware
>Intel's high end desktop chips have since then gone stale as Intel hasn't had a reason to invest any significant amount into their development
>Just buy a second hand i7 950 and motherboard with 2 x16 PCIe slots as I'm doing most of the work on GPU's anyway
>Find out about Zen
>Will probably upgrade my machine to a Zen chip when they come out if they'll have motherboards with dual full size slots
>>51434924
The latest drivers have the Fury X right at the Titan X's level at 4k.
>>51437174
Where do you say intel has gone stale? Ahead of AMD in everything and with Zen finally catching up to exactly where Intel is I don't get how Intel is the stale one.
>>51437076
What tanked AMD wasn't that they bought ATI, it was that first Intel was able to catch up without letting AMD get too much of the market by bribing OEM's not to sell machines with AMD chips (they eventually got fined for his by the EU, but only after the damage was done) and then made a bad bet on software becoming more parallel than what it actually did with Bulldozer and it's successor cores.
They created mantle to show what could be done if graphics API's were more low level and allowed for more parallelism, but it didn't work fast enough as DirectX12 became available only recently and the full Vulkan spec is expected to be finished before the end of the year. This could be the saving grace for their GPU's, but it's too late to save the Bulldozer CPU lineage.
>>51437268
They may not have gone stale in comparison to AMD, but they have gone stale in comparison to their own previous efforts. Only a 5-10% performance gain for every new microarchitecture is what I call stale.
>>51437301
Fair point. In that regard I wouldn't mind seeing AMD do well with Zen. I just always liked Intel because the way the CPUs and chipsets were designed it always seemed like Intel had it better.
>>51436815
>hurr durr
Why is /g/ so incompetent when it comes to finance?
Nvidia is trading for 3.8 times its book value
Apple is trading for 5.5 times its book value
>>51437403
>why is /g/ so incompetent when it comes to finance
($0.45) is negative 45 cents per share you dunce. That means for every share you buy you own someone an additional $0.45.
>You dunce.
>>51437076
selling their ARM-graphics-dept to qualcomm right before the big smartphone-boom wasn't really a good decision either.
>>51437951
oops. probably motivated by their large debt
>>51437916
Topest of keks.
>>51437916
Reminder that at current market value you could buy AMD and pay off all of its debt for just 4 billion USD. Actually just 3 billion considering AMD still has roughly 1 billion in cash laying around.
Even if Zen isn't a success someone will take over AMD long before they run out of money.
Intel is being fucking lazy and it might bite them in the ass in several years.
But for now it's free cash flow lmao.
>>51438301
Marketing and all that. Pretty much what nvidia does I guess. But apple is the worst.
>>51438301
Your timescale is fucked. That's not any better at all.
Testing 1234
>>51438357
Last test 4555
>>51437212
Aint Fury ton cheaper than Titan X too?
Wasn't AMD supposed to present its new driver Crimson for the video cards today?
>>51437212
Wait Fury X is at Titan X level? How
I got one of these to tide me over til the broadband is installed. Will playing counter strike or company of heroes online rape the data allowance ?
>>51438390
Almost half the price. Wish I'd bought 2 furys now instead of a 980ti.
>>51433589
Will my FX-6350 hold me over in games til late 2017 or early 2018?
>>51438450
Was thinking of building PC with 980 Ti, I guess I need to read up on Fury X now for fucksake.
Who so high TDP? Intel's old sandy bridge has better efficiency
>>51438301
That's still 50-60% better than Bulldozer/Piledriver no matter how you make the graph.
>>51438411
Whoops, I misremembered
The Fury X beats the 980 Ti at 4K and comes at the heels of the Titan X. At 1440p, it matches the 980 Ti, but lags behind the Titan X.
At 1080p, the Fury X is behind the 980 Ti.
https://www.techpowerup.com/reviews/MSI/GTX_980_Ti_Lightning/23.html
>>51438530
How does Fury X have better performance at 1440p than 980 TI, but worse in lower resolution`?
Sorry I am tech illiterate.
So Fury X in your opinion, good or bad GPU?
>>51438530
And this is without voltage overclocking the Fury X, which was just enabled two days ago with Sapphire's OC program. But I have my reasons to doubt that voltage OC'd Fury Xs will be able to beat a stock Titan X.
>>51438553
I think it has to do with overhead on the PCIE lanes. For whatever reason, all Radeon cards send a massive amount of information to the CPU through the PCIE regardless of how much work the GPU is doing, and lower resolutions tend to be more CPU intensive (the rendering load shifts from the GPU to the CPU) because of that.
Nvidia's Maxwell is designed to reduce this overhead, making them better at 1080p and lower because of reduced overhead and possibly latencies. But it doesn't scale at higher resolutions because the CPU isn't as stressed once more of the rendering is pushed onto the GPU. The CPU is able to handle the data passing from the GPU since it doesn't have to work as hard. At 4k, even a FX-8370 can keep pace with a massively overclocked 8-core i7 because most of the rendering load is on the GPU rather than the CPU.
Correct me if I'm wrong, because I only have a rudimentary understanding of CPU-GPU relationships.
>>51438653
but is Fury X good GPU to get instead of 980TI?
>>51438530
How does TechPU work
https://www.techpowerup.com/reviews/AMD/R9_Fury_X/31.html
shows Fury behind 980 TI
and
https://www.techpowerup.com/reviews/MSI/GTX_980_Ti_Lightning/23.html
shows it ahead of 980 TI in higher res
I've built planets of desktops with both Intel and AMD cpus, and I sure hope that Zenith is awesome. It's nonsense that Intel can have such tiny improvements all the time and get away with it.
That being said I expect Zen to perform somewhere between Sandy Bridge and Haswell.
>>51438715
That depends on what you're planning to do with the card. If you're gaming, what resolution are you planning to play on?
If it's 120+Hz 1080p and lower, stick with a 980 Ti.
If it's 60-120Hz 1440p, you can go either way.
If it's 30-60Hz 4K, you're better off with a Fury X.
If it's 60+Hz 4K, you're better off with two Fury Xs because of Crossfire's inherently better scaling.
If you're running compute-based programs (like predictive financial analysis or modeling multi-particle movements), a Fury X is much better than anything Nvidia has apart from its $3k+ Quadro.
I'm waiting for that dual Fiji Fury because of that and I can't fit four Titan Blacks in my motherboard.
>>51438730
>June 24
>October 27
Driver updates. AMD is slowly improving the driver performance on their GPUs. The only thing that was holding back Fiji's performance was the drivers.
>>51438801
Seems like I should get a 980 TI then. I doubt I will buy 1440p screen, they cost 2much.
What relevance does Memory Bus Width have for average PC user
AMD seems to have quite high memory bus width
>>51438801
Also, if you're doing graphic accelerated rendering, that depends on whether the program you're using is CUDA-based or OpenGL.
CUDA is Nvidia's domain and OpenGL is AMD's.
>>51438837
>What relevance does Memory Bus Width have for average PC user
Pretty much nil at your resolution. The memory bus shines at 1440p and 4K, which is why 4GB of HBM is able to keep up with 6+GB or GDDR5.
I just hope they will price them respective to their performance, I only play WoW from 2010 pretty much and have a workstation do my rendering (xeon). Does anyone remember what the initial 9590 price was, wasn't it close/past 500 dollars?
>>51438837
>wants to buy $650 GPU
>can't afford decent monitor
>>51438890
>Does anyone remember what the initial 9590 price was, wasn't it close/past 500 dollars?
That was OEM/bulk prices. AMD never planned on releasing CPU-only 9590 because they thought consumers wouldn't be able to cool their CPUs and blame them for chip failures. This was compounded by the fact that most AM3+ AMD boards are not compatible with the 9590 due to its insanely high power requirements and would literally catch fire if you tried to run them at stock voltages.
>>51438837
>Spending ~$500 on a GPU
>Not spending at least $250 on the screen to use with it
I'll never understand you people.
>>51438553
>>51438715
>>51438553
Fury X is more futureprove than 980ti because it actualy supports dx12, Nvidia did some fuckery again and doesnt really support dx12 on hardware level.
>Powerdraw
Fury X consumes more Power than 980ti, but to be honest if you can afford highend you shouldnt care about 5-8€~ per year.
Also do you use Tripple A+ freezer, hairdryer etc. and have a platinum+ certifiyed PSU? if yes ok buy nvidia, if not you dont really care about a bit more powerdraw and you can ignore that, which is fine because basicly nobody does and people who cry about muh amd powerdraw are big hypocrite fanboys.
>much amd no driver meme
amd drivers are really good nowadays.
>muh housefire meme
actualy the meme came from nvidia cards 4xx series who actualy caused some fires and nvidiafanboys tryed to coin the meme on the 290x cards because they were only available with stock coolers for a long time and like we all know stockcoolers suck huge amounts of dick doesnt matter if you use it with nvidia or amd.
In todays time aslong you buy GPUs with costum coolers on it for example from EVGA/ASUS(Nvidia) or Sapphire/MSI(AMD) will run cool as ice and you never gonna have any probs with temps with either brand.
>FuryX vs 980ti
if you only use 1080p and that for the next 1-2yrs dont buy either, what you are looking for is a 980 or Fury(nonX)/390/x ignore 970 because its a memecard(3.5gb which is not enough for todays games and not future proof).
1440p/4k
If you want to keep your card for at least 2-3years buy AMD because they are more futureproof (dx12 support and they dont gimp drivers for "older" cards like nvidia does instead they actualy improve).
If you only want to keep your card for about a year until amd and nvidia release thier new 14nm Cards you may buy 980ti EVGA because the futureproof doesnt matter and atm they are ahead in some games with about 1-4% basicly 2-5fps more if you really care about that.
>october 2016
we can make it, just have faith
>>51438837
>I doubt I will buy 1440p screen, they cost 2much
nigger you dumb 600+$ into a gpu but dont have 250~$ for a 1440p screen?? you might as well buy a 980 or 390/x
>>51438942
You forget to mention the coil whine most Fury X exhibits at load (it was never fully addressed even now), how load the Fury X is at idle (although it doesn't get any louder at load), and that Fury X is not compatible with many low-end cases if the CPU is already being water-cooled with an AIO. Only now are we seeing cases under $70 that have more than one places to mount a radiator.
>>51438975
>>51438926
I have 144hz 1080p.
what' are the recommended 144hz 1440ps screens then?
Is techpowerup reliable site for objective reviews
>>51438995
There's only a handful, and at least three are G-Sync compatible. They're usually not worth their retail prices with their build quality, though.
I'm waiting for a decent UW 1440p 144Hz monitor to upgrade from my three 144Hz 1080p monitors.
>>51433763
i hope they are right
>>51439022
144Hz gsync is a meme. You can't see tearing at 144Hz.
>>51439022
So if I am not swimming in money I should stick with my 144hz 1080p and get a slightly less powerful GPU (basic 980)?
Thenks.
>>51439037
b-b-but muh Linustechdicks...
>>51438991
I work in a techstore and we rarely had problems with fury/x coilwhine only with the old pump noise level which is fixed.
Most coil whine RMAs are 970/980 (evga seems to have the least coilwhine here) and 980ti from Gigabytes
>>51438995
Asus or BenQ
>>51433772
these amd "engineers" are so dumb, they could reduce that design size by a lot by just reducing the arrows size, the margins on each component and even the blank space between them. the schedulers didnt need to be made bigger either since they will be doing the same thing, and i dont know why they give up so much room for a decoder when most users dont even deal with encryption
fucking dumbasses man, kinda shows how pointless college is
>>51439060
TN or IPS from Asus?
Got Asus MG278Q 27" LED TN and Asus MG279Q 27" LED IPS for 600 both in my local areae
>>51439038
Get a vanilla Fury then. It's 90-95% of the performance of the Fury X at all resolutions.
>>51439060
>old pump noise level which is fixed
I've seen newer Fury Xs (the ones with the newer Cooler Master block designs) in a test bench that sounded like my old Corsair H100i at full tilt. At idle.
I'm getting my dual Fury Xs/Geminis/whatever, but it's going on a custom water cooling loop for that reason alone.
Also, the coil whine seems to be directly proportional to the GPU load. It gets louder the harder its being worked. It's something AMD was never able to fix, even with better quality capacitors and MOSFETs.
>>51436395
Shit, they are using their hbm, that will be insane
>>51439095
>local area
buy from websites... I dont know what you burgers use but I see newegg trowen around many times sorry but im EU
Are Zen's cores 'cores' or actual cores? Are they going to get in trouble again?
>>51439160
If you couldn't tell from the diagrams, they are full-fledged cores.
I reckon that they're still weaker that Haswell, but only just.
>>51439099
>I´ve seen newer....
Full Costum loop are always best for cooling, noise level depends on the parts you use for it.
Like I said I normaly dont build test bench systems only "normal" cases and most of the time I only hear the Fans with acceptable noise level.
Have to say I don´t have the pcs next to me but under the table, because I find it autistic to have it on my table. So its fine for me can´t say how noise levels are if its next to you ON the table.
>>51439172
Also Zen will have SMT.
>>51439185
I'm doing it just for the noise level and better VRM cooling.
As for the test bench, it was a fully open test case so it was louder than it would be in most cases. But, it was still as loud as my old Corsair that I used on the test bench. To its credit, it does not get louder when running GPU stressers. I don't think the fans even ramp up, if they do it's probably up by just under 500 rpm.
In a closed case with sound deadeners, I'd imagine that the Fury X would still be unusually loud compared to the rest of the computer.
But that coil whine...
>>51439246
>But that coil whine...
You maybe had a bad egg.
We have 1/20~ Furyx get rma
We have 1/12~ 970 rma for coilwhine
We have 1/15~ 980 rma
We have had 9/20 980ti rma for coilwhine, after we banned gigabyte from our shop its now basicly the same as furyx 1 of 20~ 980ti(evga being the least of the rma´d) but it always depends we also had a case were we had rma a single 980ti 3x times befor we got one without coilwhine (gigabyte btw)
>>51438801
>a Fury X is much better than anything Nvidia has apart from its $3k+ Quadro.
False because there's no software for opencl as opposed to cuda, unfortunately.
>>51438864
>doesn't know the difference between opencl, opengl, directx and cuda
>tries to give advice
Warota.
>>51439408
What is the difference then? And why are Nvidia cards inherently better for film makers and profession media? You should know this, Mr. "Expert"
>>51439398
>False because there's no software for opencl
False in the financial world because there are quite a few software that built around Opencl.
>>51433621
8 core Summit Ridge absolutely will not be $300.
AMD is done being the budget brand. Expect a $500 i7E competitor.
Expect the flagship Raven Ridge APU to be priced around $250 to compete with intel's Iris Pro equipped desktop chips.
>>51433589
>>51433620
to wait and see, if all of this happens i might finally buy an AMD CPU again
>>51439589
amd highend for 300$ the dream gonna be instabuy if its really around 300shekel
>>51437022
have fun waiting a whole year for your upgrade, goy!
>>51439589
I will take a 4 core Zen for i5 prices if they put hyperthreading on it.
>>51439589
[citation needed]
Zen cores can't be that big considering they are working on 16 and 32 core APUs for supercomputers.
The 8 core desktop chip could be pretty small and cost efficient.
What are main differences of AMD and intel CPUs
and what are the main pros and cons in comparison
>>51439722
Watch it be $300 for "muh price parity". Fucking Suu wants AMD to move away from the budget choice image.
>>51439722
Remains to be seen how they'll handle binning and segmentation of their SKUs, but I imagine there will be just that.
>>51439727
Lisa Su came out and explicitly stated that AMD would no longer be the budget brand. They did not sell the FX 8350 for $200~ because they wanted to. They do not have shitty margins on most of their chips because they want to. They did that because they had to, because they couldn't compete against intel if their chips were priced similarly.
They're not going to sell high performance chips for a fraction of the price of intel's any more.
>>51439087
>>51439756
So their non-budget CPUs will perform at the level of Intel's 2+ year old (at release) architecture.
Goodnight sweet prince (pic related)
>>51439756
if that is true and zen doesnt perform at least at the same level as Intel, amd can fuck off.
I buy amd because of price/performance if amd want to compete with higher prices they have to be at least intel level or my next cpu is intel because better price/performance
Wait ©
Futureprove™
>>51439878
AMD would be shooting themselves in the foot if they don't release a 8-core non-hyperthreaded version or 4-core hyperthreaded for under $300.
But Intel would be the one doing the shooting if Kaby Lake drops its price compared to its Skylake equivalents by just $50.
>>51439087
kek'd
>>51433877
TICK TOCK ON THE CLOCK BUT THE PARTY WON'T STOP
Going to buy Fury X
How should I prepare my anus for the coil whine
>>51439979
>fall in the 4GB HBMv1 meme
>>51439834
Ultimate performance not only comes down to architectural nuances and IPC, but also what clocks they can achieve. Their quad core chips could clock high enough to cover any performance disparity vs a similarly priced i5.
They're not going to market any of their chips in an illogical way.
How it should be
>4-core APU (No HT)
$100-170
Competes with Intel i3 and i5 S and T
>4-core APU (HT)
$200-270
Competes with Intel i5 and i7 S and T
>8-core CPU (HT)
$300-500
Competes with LGA 2011v3
How it's probably going to be
>28nm Excavator APU
$100-150
Competes with Pentium and i3
>4-core APU (No HT)
$200-270
Competes with the i5
>4-core APU (with HT)
$300-350
Competes with the LGA 1151 i7
>8-core CPU (HT)
$500-1000
Competes with the LGA 2011v2 i7
I can sense Zen pricing itself out of the market with ambitious pricing.
>>51440027
What the fuck is wrong with ASsassins Creed
>>51439915
>>51440043
>4-core APU (No HT) Competes with Intel i3 and i5 S and T
They wouldn't waste die space for that and make 2 cores 4 threads and 3 cores 6 threads. I expect 3 core one at the sweet spot for price per performance.
>>51439888
TRIPS NEVER LIE
thread/
>>51440102
AMD doesn't have the money to create numerous dies like intel does.
There will be one master Summit Ridge die which is what Opteron MCMs will also use, and one master Raven Ridge die. All subsequent SKUs will be binnings of these.
Intel can make 4 different desktop dies, and 4 different mobile dies. AMD can't.
>>51440102
>2 cores 4 threads and 3 cores 6 threads
That's not how Zen is structured. Each module comes with either 4-core CPU or an R7 GPU. There is no smaller module than that. There is a possibility that they can take lower-binned CPU modules and disable some cores, but I think that's possible based on how the cores are integrated with each other.
>>51440146
>>51440147
OK than, no possibility for cheap Zen stuff for real. I hope they make some godavari for AM4 for budget.
>>51440187
>I hope they make some godavari for AM4 for budget.
That's what that Excavator APU is for. It will be the first chip to come out with AM4+ this coming Spring. HSA enabled, but HSA hasn't picked up any steam since AMD's been hyping it up all of last year.
>>51433877
except Intel's clock broke and is now going tick tock tock tick
>>51440187
Binned chips are always cheaper, anon. Not all of their chips are going to be priced at the top of the market, pricing structure will still work the same way.
Bristol Ridge on AM4 is the desktop version of Carrizo, it'll probably launch around $130 since they're cheaper to produce than Kaveri/Godavari. Generic 28nm HPP instead of custom SHP.
>>51433589
All they need is to BTFO this i5 and I'm happy. If their 8 core behemoth is close to the 5960x/6930k, then I'm all for it. In fact, they can come close to the 5820k and I'm all for it. I want to use muh emulators and have the extra cores for rendering and a comfortable VM.
Also, their Greenland better deliver. I have about $3k saved for an entire build and would love to go full AMD.
>>51440147
i want my Zen without a fucking gpu why would i ever want a gpu in my cpu?? thats fucking trashtier. I want more power not more bloat hardware
>>51440259
Summit Ridge is CPU only.
Raven Ridge is the APU, though APUs are now "mainstream."
>>51440048
The result of hundreds of people working on the same code together.
>>51433850
Phenom B55 here running at 3.8Ghz stable
>>51440259
The GPU component is only for APUs and HSA-enabled enterprise CPUs.
It goes like this
Consumer CPUs:
>1 CPU+1 GPU = APU
>1 CPU+1 CPU = CPU
Enterprise CPUs:
>1 CPU+1 APU+1 Custom+1 Custom = SOC 1
>1 CPU+1 CPU+1APU+1 Custom = SOC 2
>1 CPU+1 CPU+1 CPU+1 APU = Enterprise APU
etc.
>>51440317
I don't think Basilisk is not going to be sold outside of pre-soldered boards, so it's not relevant to the Zen we're going to come to know.
>>51435047
No. No we are not.
The Fury X is about 2% better than the Titan X. The Fury cant beat a reference 980 or some of the better superclocked 970s.
>>51440317
This is 100% fake, all of those leaks are.
Bristol Ridge is not Zen based, it is just repackaged and tweaked Carrizo. It uses Excavator modules.
There is no Basilisk chip. Basilisk is a really old code name for an APU that was scheduled to release two cycles after Kaveri. Construction core development ended, and Basilisk never got out of the concept phase.
The name itself came from a Linkdin profile. That was the only mention of it anywhere, and it was deleted immediately after it was found and spread around S|A by that schizo ponyfag Seronx.
AMD's low power APU to replace Kabini/Beema is called Stoney Ridge. Its a singe Excavator module and 3 GCN CU.
>>51440343
Its not real.
>>51440317
estimated release when?
>>51440370
So where is the info from that Summit Ridge is going to be 8 cores?
>>51440317
>>51440343
>Up to "x cores"
The language they're using doesn't imply hard numbers. As one anon said there's probably four cores per module, but from what we've seen they may not be linked in pairs like bulldozer. Either way I suspect we'll see 6/8 core and a super low end 4/8 core from summit ridge.
I mean these days there's always a push for lower power draw parts and a good way to get that is to cut cores off anyway.
>>51440397
Sweclockers at the beginning of this year released a bunch of details on AMD's upcoming Zen based chips. They were the first to use the codename Summit Ridge, and they stated it would be 8 cores and have a 95w TDP.
http://www.sweclockers.com/nyhet/19954-amd-zen-forst-ut-i-summit-ridge-for-14-nanometer
Actual roadmap slides that were under NDA got leaked, and some patch notes later listed details for Zenver1 and Stoney.
>>51440400
dont know why they dont just make a 2-store cpu and put 20 cores in it
>>51440483
pic also related
>>51434911
Its not 2003 anymore m8.
>>51438426
Hahaha yes
>>51440370
>not real
Should have figured since Zen was only stated to have 4 cores at its smallest configuration.
Plus it's still using FM3 as the socket, when it's been confirmed/leaked that it is officially "AM4"
>>51438301
>>51440814
Have something actually relevant.
Zen is nearly 80% higher IPC than Bulldozer.
Excavator is almost 28% higher IPC than Bulldozer.
The only Steamroller and Excavator chips we can test are lacking an L3, while the Bulldozer and Piledriver chips have one. If the APUs touted an L3 then their performance would be even higher, which would push the projected figures for Zen up slightly.
>>51440930
FM3 has been the go to internal development name for new sockets at AMD for a while. Before Kaveri launched some Coreboot files started listing socket FM3. There are references that go back before that too.
I think that Seronx guy is the one who made the fake slides and "leaked" them to that german forum right before AMD's financial analyst day.
>>51438301
>hurr every chart scale has to start from 0
it's still better than Apples "2x performance!" with bars' values being "1" and "2"
>>51438301
>>51441073
Isn't it mostly going by years on AMD's chart?
>>51438463
Buy regular fury. Sapphire or asus make the best coolers.
>october 2016
Godamn AMD.
>>51441393
By that point Intel will have released an i3 that can beat it.
Why does Fury X use so much power compared to 980 TI or Titan X
>>51441479
It doesn't even use that much power compared to those two. Besides, it idles well on multi monitor and higher refresh monitors while the others panic the fuck out with a higher refresh monitor.
>>51441511
Is there comparison of Titan X - Titan 980 GTX Ti and Fury X in 144hz monitor
>>51441479
It doesn't.
>>51441523
https://www.techpowerup.com/reviews/AMD/R9_Fury_X/29.html
it does tho
>>51441588
>980ti uses half the power fury x does at same performance
lel
>>51441588
OH NO 20W MORE ON AVERAGE
THIS WILL SURE BANKRUPT US ALL
>>51441522
For the 980Ti
http://www.pcper.com/news/Graphics-Cards/Testing-GPU-Power-Draw-Increased-Refresh-Rates-using-ASUS-PG279Q
>that jump from 120Hz to 144hz
>>51441632
>literally 200% the power consumption of 980TI let alone how fucking hard Fury X shits its bed under stress
????++?? the card A SHIt
>>51441588
No, no it does not. The pic I just posted even tells you whats wrong with TPU's, and every other site's reviews.
Tom's started using a multi probe oscilloscope set up and measuring actual line power delivered to the components. They look at power consumption in real time, and log it.
TPU and other places use laughably in accurate killawatts, or they try to estimate power consumption. This at best will let you see momentation peaks which are not at all indicative of the average power consumption.
Fury X will peak above 400w for micro seconds, the average load consumption is still well within 250w. Average load it draws slightly less than a reference 980 ti.
Educate yourself, shitposter.
>AMD
DOA
>>51441663
Why is your test vouching for AMD being better 'better and more honest'
Seems biased senpai
Just going to overclock my i5 4960k and sit in it while I wait for 8 thread Zen cpus. Zen is the only reason Ive uet to upgrade to skylake.
Please deliver AMD. It caised my pain to move away from my FX piledriver but it was just too hot and loud for what I needed it for.
>>51441663
>Tom's
isnt this the site that rated price as 'higher = better'
>>51441710
Because they used sound methodology and the correct instrumentation to measure power.
It is scientific, and concise. A wall wart measuring power is not.
Shitposter.
>>51441746
Sorry
>>51441656
>>51441588
You're just fucking retarded
>>51441768
BUt that's worse than 980 ti
>>51441710
because toms hardware is basicly unbiased if all they shill a little for intel but gpu whise they are as fair as it gets.
>>51441712
hot/loud sound like a cooler problem to me not a cpu problem.
Never had temp problems with my amd cpus nor intel cpus and I always use the macho cooler
>>51441804
What the shit are you trying to prove you retarded faggot? If you can't afford ~30W then you shouldn't be buying any of these cards in the first place.
The Fury is a vastly more powerful GPU for compute tasks and has far more stream processor, which is one of the reasons that it draws more power too.
>>51441824
Thanks for this.
So how do you deal with coil whine with Fury X? Isnt that issue with it
>>51441816
are you 12?
back to elementary class for shilling
>>51441816
But it's THIRTY FUCKING WATTS.
If you can't afford to pay for fucking 30W then you're seriously so poor that you belong in Africa.
>>51441804
but there it is spending too much power agian
>>51441865
30 watts is still 30 watts.
>>51441865
But the point is I can get 980 TI that does same with less power consumption
Why are you being retarded? Why wouldn't I want more efficient and better tech?
Why is it bad to want it? Both are even in same price range
>>51441881
You spend more on your fucking cooling system than that.
>>51441891
And? Why should one just not care?
>>51441891
I have 7 fans each draws 400mA each. 0.4x7=2.8 amps x 12v =33.6 watts.
Are you spying on me?
>>51441884
>>51441929
But the 980 Ti doesn't do the same. That's only the case for vidya (granted you're 12 and that's your only use case). It depends on the workload and the clear winner when it comes to GPGPU is the Fury X in this case. For video game graphics it's most likely AMD's driver overhead which slows it down (this is pretty much proven at this point).
I do care about efficiency, but 30W is not significant at all. It costs me literally less than $2 to pay for that shit, and it doesn't significantly load the PSU any more either.
>>51441884
so m8 you only buy stuff with the lowest power consumption.
Please post me your Platinum/Titan+ certified PSU, your AAA+ energysaving TV, your AAA+ fridge and please tell me you only use lower power consumtion light. You don´t awww little hypocrite fuck off fanboy.
Nobody gives a shit about theoretical 10$ more in the powerbill
>>51441947
Yes, anon.
Do you remember when you suspected someone was under your bed?
That was me.
>>51441816
35W you faggot. My shit third world island's energy costs are $0.2 per kWh. Assuming you game 8 hours every day, you would be looking at 2920 hours per year. Thus, your energy costs would be ($0.2 / 1000Wh) * (2920h) * (35W) $20 every year. If you live in a country that isn't shit, you would be looking at much less than $20 a year.
>>51441968
What is wrong with you? Why shouldnt I care about power?
>>51441834
When you're running an fx-9370 at 5.0ghz it gets hot and loud. Temps were under 62C and cpu was stable, but it only gave 60% of the gaming performance of my stock i5.
Only thing it did Better was video encoding and heating my room.
>>51441979
But 980TI is more effiient unless you farm bitcoins.
>>51441983
because you dont, if yes post me all your power saving stuff also you always cut off tech and dont let it in standby like tv because that cost you money and wastes power. So we see you are just hypocrite and dont really give a fuck about power consumption
>>51441991
60% against stock i5, WEHW LAD nice try
>>51442006
Why am I hypocrite if I care about power saving?
You have some issues, damn son
>>51442020
LIKE I SAID YOU DONT CARE ABOUT POWER SAVING, there is so much you could really save power AND $$$ instead of whining about fucking 30watts.
Like I said you don´t have powersaving tech at all, you let devices run in standby (just that costs a normal house about 90-100$ per anno), not using power saving lightbubs. You just want the hold the MUH amd comsumption meme up so you have something to shill about.
At least try shill please.
>>51442006
I always switch everything off at the plug when not in use and I have a smart meter to help me track down vampire power drains.
>>51442042
But I am saving.
When I am choosing the more efficient GPU.
This is caring about power use.
>>51442019
That was just a rough number i tossed out there. But in games like the new wolfenstien I was getting fps dips down to the teens on my 9370. This is paired with an r9 390.same driver same settings same card but using an out of the box i5, I was getting dips down to low 40s at worst.
This all being at 1440p.
Not an intel fanboy by any means. I want AMD to make a comeback so i can ditch intel asap.
>>51442005
A9X is has the best performance per watt unless you care about legacy x86 programs. Where is my K12 for desktop ARM, AMD?
>>51441845
>has far more stream processor
Yeah it's sad, AMD have no drivers for all his 4096 sp and can't beat 3072/2816 nv core
>>51442076
>unless you care about programs 99% of users care about
? !?!?!+
>>51442110
Back to /v/ with your gamen shit.
>>51442125
>Only games are x86
u fucking wot?
>>51434400
Except it does. There's a few Z170 mobos out there that have DDR3 support.
>>51441804
>dat 390x
>>51442155
No it is: only "gamers" think 99% care about x86
In reality 99% of the users only care about their smart phone.
>>51442210
hue
>>51442051
>paying $100 more for gpu
>look guise I save $20/year!
>>51441712
>4690k
Well, a 4-core Zen with Hyperthreading may just edge out the i5-6600k
i5-6600k @ 3.5GHz/3.9GHz
>Passmark CPUMark: 7739
A10-7800 @ 3.5GHz/3.9GHz
>Passmark CPUMark: 5205
Assuming part of the IPC increase is due to hyperthreading (which it most likely is given what we know about the architecture) and assuming that IPC increases is directly proportional to a CPU's Passmark score (which it most likely isn't, but fuck it why not)
Zen 4-core APU @ 3.5GHz/3.9GHz
>Passmark CPUMark: 8005
But, I highly doubt that it would be this low with hyperthreading, since on an Intel processor hyperthreading boosts the score by 30-35%.
Who knows, maybe Zen has a higher IPC without counting hyperthreading, which makes it potentially even with Skylake.
>>51435366
I bought my 6300 in 2013. Due to the fluctuating value of my currency I cannot buy anything better for the price I paid back then. I hope that USD will be cheaper in 2016.
Everyone, please take a moment to enjoy this lesson in obvious trolling.
Guide for idiots: Odds are the troll(s), evens are the victim(s).
>>51441479
>>51441523
>>51441588
>>51441768
>>51441816
>>51441865
>>51441884
>>51441968
>>51441983
>>51442006
>>51442020
>>51442042
>>51442051
>>51442389
I love you /g/uys :^)
>>51442488
>Using more power does not mean using more power
Excuse me
>>51442502
HE DID IT AGAIN! HAHA!
>>51442439
Hoping so. I do video encoding and 4 threads just isnt enough. Takes too long. Was eyeing the i7-6700k but waiting to see what Zen ofders for 8 thread. Dont wamt to buy a $330 chip to find it gets stomped by zen $250 chip.
Not to mention Z170 only has 16 pcie 3.0 lanes where as am3+ had 40 pcie 2.0. Hoping amd doesnt do away with that.
>>51442164
Except it DOESN'T. Intel explicitly stated DDR3 will burn out Skylake's memory controller because of the higher voltage, eventually killing it.
DDR3L will work fine though.
>>51442523
>40 pcie 2.0
God, I missed having so many "fast" lanes. But I doubt we'll be seeing 40 lanes for an AMD APU. They're probably disable a few lanes if you put in an APU like with the 5820k over the 5930k. They're sharing the same socket, so it only makes sense that they limit the number of PCIE 3.0 lanes for CPUs that can't handle the bandwidth.
We'll have to wait until the first boards come out Spring of next year. Merging the APU and FX socket could be a blessing or a curse (like giving us no more than 16 PCIE 3.0 lanes for BOTH CPUs).
>>51442488
>summerfag
Implying phonefags aren't the real cancer here
>>51442569
Ugh i know what you mean. Hopefully amd boards have all the little things that intel boards have as well.
Even my $140 asus z97-ar has an m.2 socket, supports nvme, and has an intel LAN chip. Not even sabertooth or ROG amd boards offer those.
>>51437301
5-10%
More like 2% faster and using 5% less energy.
>>51442716
>Not even sabertooth or ROG amd boards offer those.
Each of those features on the Z97 and Z170 boards are aftermarket controllers from ASMedia or Intel. I think it's left out because there is no financial benefit to adapting them to AM3+. It's up to them to make the firmware work with any board they're put in. I know Killer offers their NIC on at least one AMD A88X board and ASMedia put their M.2 controller and USB 3.1 controller on two AM3+ boards.
>>51438301
Nice scale on that x-axis you fucking retard
>>51439245
Disgusting.
I'm going to miss cmt...
>>51442755
Sucks that their a88x and athlon didnt take off. If the athlon had more single thread power, it would have been a major contendor against Intel's i5 at a much lower price point.
Is the 7870k faster than the 860k even though it has an igpu? The a88x motherboards are actually pretty solid.
>>51442825
>Is the 7870k faster than the 860k even though it has an igpu?
They're the same chip. The A10 7870K is the highest binning of the line, and unlike the Kaveri based chips it has a soldered IHS for better temps.
You'll be able to hit equal clocks with less voltage with the Godavari A10.
>>51440027
Could you have picked a less optimised game?
Also.
>that file name
Not even sure if pic is real...
>>51442860
>>51442825
But FM2+ boards are limited in how much you can overclock them via the BCLK. I think you'd have much better luck hitting higher frequencies using an X4 760k on an A85X FM2 board.
Daily reminder that you have a moral obligation to by Zen and support AMD since Intel an Nvidia are trying to destroy the open market
>>51442882
People have hit 5ghz with Kaveri, many more hit 4.7ghz.
Its rare to get one, but they're not at all limited in clocks by the mobo at all. Unless you were talking about LN2 max theoretical OC, but no one gives a shit about that.
>>51442882
Why would you use blck if you can just tap the voltage and multiplier?
>>51442860
I figured as much. Still tempting though. How do you think a 7870k paired with a 390 would fair at 1440p?
>>51434264
Last time I checked Skylake barely gets better than haswell performance
>>51442904
>People have hit 5ghz with Kaveri
Those are really good chips. I think 4.6GHz is what most examples can hit.
>Unless you were talking about LN2
No, there is a limitation within practical overclocking. Most 760k owners who've tried overclocking on A88X boards have not been able to reach the same clocks their X4s have been able to hit on older FM2 boards. It's suspected that its either because of a power delivery change on FM2+ boards or that BCLK ceiling.
Is Kaby Lake a tick or a tock?
>>51442997
Neither, it's a refresh like Devil's Canyon was to Haswell
Tick tock is dead along with Moore's Law
How does Fury X overclock for gaming
I know Titan X overclocks like a fucking miracle
Why do niggers keep posting about Fury X and Titan X
This isn't a GPU thread
>>51443044
both have Xs in their name..
>>51443033
>How does Fury X overclock for gaming
Like shit, which is funny because they said it would be an overclockers dream even though the memory is locked
>>51443063
Does it make them go faster
>>51443073
Guess I get some nvidshit then
>>51434911
More like 40% of the performance for 60% of the cost.
>>51440979
It' a shame because (in full threaded work) the 8320 was matching and beating 6 core Ivy Bridge chips.
Excavator could have been a match for high end Haswell, but AMD no monies.
>>51443122
More like 80% the performance for 100% the cost if it's true they're done with being the budget company.
>>51435450
I'm hoping my shitty Asscock motherboard can survive until Zen hits. Sometimes the system will hard lock up even when absolutely nothing is wrong on the OS layer and requires a power cycle. Other than that I love my 8350 build.
>>51443228
AMD ended up, very slowly, fixing tons of the architectural issues in the Bulldozer family. Steamroller actually would have had huge IPC gains if it weren't for cache latency increasing, Excavator improved on this a bit, but its still not competitive with intel's.
If they had just had more time put into development before the rush release of Bulldozer we probably would have seen something like Piledriver-Steamroller in 2011. Of course they'd still be limited by GloFo's 32nm node, but that was out of their control.
Makes one wonder what could have been. If a 14nm 4 module Excavator based chip had Zen's new low latency/high bandwidth cache system it would be quite something. The construction core family probably would have continued development. Just one update to improve serial integer performance and they'd have had a real winner.
Goes to show just how debilitating bad management is to a company.
>>51433589
>40% more IPC
So it's still behind Intel by like 200%?
>>51443417
>look mom I'm late to the troll party
>>51443486
>look mom i'm a butthurt amddrone
>>51443417
40% IPC would put it about equal to haswell
>>51443497
>implying
>>51443514
>equal to Haswell
And Kaby Lake will be ~15-20% above Haswell's IPC. So Zen will be 15-20% behind its competition.
Will update once zen launches, not sure what to do with it though
>>51443570
>And Kaby Lake will be ~15-20% above Haswell's IPC
Not really no, skylake was like 5% faster, and kaby lake is mostly just going to bring native USB 3.1 support
>>51443601
>kaby lake is mostly just going to bring native USB 3.1 support
No, Kaby Lake will have the prerequisite IPC increase like Broadwell over Haswell.
>>51443570
20% lower IPC
100% more cores
Kaby Lake will be destroyed
>>51443622
Where are you getting that? broadwell is hardly faster than haswell, same way skylake is hardly faster
The major gain broadwell has is in it's iGPU but no one cares about that and what they had to do to get their iGPU that fast raised the price of the chip substantially
http://www.pcworld.com/article/2940489/the-truth-about-intels-broadwell-vs-haswell-cpu.html
>>51443694
>igpu
It will matter in dx12.
>>51443727
it really won't, mostly because no one has the chip, and a GPU uses different RAM, which would make syncing it difficult
>October 2016
So once again AMD giving too little too late.
>>51443727
small, bare-bones iGPU setups won't be relevant to any new features for another 2-3 years at the least.
APUs are a different story, but those pathetic yet feature rich intel GPUs will do nothing but choke on DX12 functionality.
>>51443760
Cannonlake is delayed. Intel has nothing worthwhile in 2016
>>51433589
I'm worried. They want to be the enthusiast brand, but it really is coming out late.
I don't know if they're all SMT enabled, so I'm just assuming they are.
If they release a quad core for $300, then it better at least be on par with an i5.
Also, a $500 octa core is dangerous because not only is the 6930k going to be 8 cores, it still has to beat the $400 six core 5820k/6820k.
I'm not worried of it beating my 3570k so I can finally upgrade, but I'm worried for the higher end.
>>51439366
you mean gigashite, amirite
>>51443570
>Kaby Lake will be ~15-20% above Haswell's IPC
Are you retarded?
Kaby Lake is not a different architecture. Kaby Lake is to Skylake what Haswell Refresh was to Haswell. You're only getting a new chipset with whatever new features are out by then, and a 100 MHz increase in frequency for the same price.
>>51443958
They were an enthusiast brand before, no reason why it can't be so again.
As far as segmentation goes, some chips will probably have SMT enabled, others won't. They'll do everything possible to fill the gamut from Pentium competitor to i7E competitor.
The real question is why the fuck did AMD stick with the Phenom II architecture until now? I honestly don't understand what the fuck they did for the past seven years. They just kept releasing the same core design over and over again... they're no better than the jews at intel or nvidia. And another year now? They'll need to price Zen so aggressively they're bound to lose a shit ton of money, and even then it might flop hard. The whole industry was so much better 10-15 years ago, now it's just a matter of choosing whose kekold you'll be.
>>51443958
They want to be enterprise brand more than anything. The real high profit is on the server segment.
>>51444036
Considering how pointless opteron chips are for servers they'd have to do something crazy to be viable there
>>51444054
They made it really clear that the server market was their biggest growth market. They're expecting more revenue to come from there than anywhere else.
>>51444036
>2003+12
>AMD
>servers
kek
well meme'd
>>51444069
Well I guess anything is going to be better than opteron, mostly they're going to have to sell people on the ARM on die for security portion
>>51444054
>they'd have to do something crazy to be viable there
Yeah, definitely! Something crazy like maybe match the IPC performance of the Haswell/Skylake cores Intel uses on the highly popular and profitable Xeon line.
Is there any indication they are doing something like that? Like maybe developing a new architecture to replace Bulldozer or something, right?
>>51444080
Yeah they designed Zen for gaymur guys.
>>51444115
They actually did design it partly for the gamer crowd. Desktop sales are tanking because of tablets and laptops, but the only thing fighting that trend is the gaming market.
Why do you think Skylake launched with unlocked versions first and Broadlake being passed over despite having a broader appeal to most home users?
>>51444115
I hate Intel too and hope Zen will be good, but, come the fuck on, they haven't released anything good in years. All those shitty graphs don't mean anything,
>>51444150
Intel is profitable as fuck, AMD is on the verge of bankruptcy. Trying to cater to Intel/nvidia fanboy market is beyond retarded if you consider the ever falling radeon sales.
>>51444157
They made PS4 and Xbone chips, I hear they are popular. Too bad it doesn't do shit money.
>>51444098
Crazy as in do something new for the server market
>>51444098
AMD isn't going to match Haswell with Zen. They might be within 5% or 10% in most integer ops, but they've got no chance of competing on the FPU side of things.
What they have to offer, which they've already hinted at, is "disruptive" bandwidth. They're going to have better socket to socket, and port to port communication than intel. SeaMicro's Freedom Fabric was the only worthwhile thing the company had to offer, and AMD kept that IP. They kept it, and they went to work building upon it.
That means their 5-10% lower performance will matter less and less the more sockets you add to a cluster, until they come out on top. 20 quad socket racks with Zen Opterons could start edging out 20 quad socket Xeons. The more sockets you add, the greater the advantage AMD offers over the competition.
Price them to undercut the Xeons and bam. You have a winning product.
Everyone with old hardware in their data center will be looking at AMD when they upgrade.
>>51444254
They don't need to do something new to be viable. They just need to offer a product that is as good or better than the competition, that makes them viable.
>>51444433
>AMD isn't going to match Haswell with Zen. They might be within 5% or 10% in most integer ops
We'll have to wait and see that.
But true, they don't even have to. If they have a chip that is 90% to 95% as fast as a Haswell chip but costs less than 90% as much, they already have better value.
On top of the other stuff you just mentioned, of course.
>>51444491
Which means they need something truly new, not just an IPC lift, because an equivalent xeon is still going to beat them
>>51444570
They are talking about APU opterons for HSA and GPGPU loads. As for something truly new, they have K12 trump card but low power dense ARM servers aren't getting the traction they forecasted so it is delayed.
>>51444570
Do you know how to read?
If a Zen CPU provides better value, that means it's faster than the equivalent Xeon at the same price range. In that case it doesn't match the flagship Xeon by 5% to 10%, but flagships are not the most important product in their lines. Specially not for Intel, where they flagships have a very disproportionate price increase over the other models.
You're an idiot if you think every company just goes "just buy whatever is the absolutely fastest there is out there, regardless of how much it costs, to hell with cost-vs-benefit!"
>>51444663
Performance for watt is much important for servers. If they can deliver that, they will have a chance.
>>51444663
But it won't be faster, at best it will match haswell/skylake, and if it costs less then yes people will buy it, but only if the performance is really up to par, right now you can get a 4c/8t Xeon for the price of an 8 core opteron, making opterons pointless
>>51444710
Why not a 16 core opteron at 1.4ghz or some low frequency to make it use less energy.
Should rape the quad core Intel xeon?
>>51444744
Transistor libraries don't work that way.
When the chip is fabbed the vendor will pick a certain target range of voltage and clocks, and the specific characteristics of the transistors themselves are altered to reach this point.
If the library chosen has its sweet spot at .9v and 2.5ghz it only has a little bit of wiggle room on either side of them. You can't just lower clocks forever and see significant lowering of power consumption. Nor can you continually lower voltage. The chip will simply cease to function.
>>51444744
Only if the workload is running tons of virtual machines which is just about the only thing opterons are good for
Really what would save them for the time being is releasing dual AM3+ motherboards with ECC support
>>51444805
This. It's why underclocking some CPUs causes instability and why undervolting only works on a very specific number of CPUs.
>>51444805
So then why not design one? Can't they modify an existing design?
I don't really know much about this, but going with more weaker cores that'll consume less power seems like a good idea?
>>51444813
Do dual am3+ or ecc?
What the hell are they thinking?
>>51444945
>So then why not design one? Can't they modify an existing design?
They could, its just a matter of having the fab produce a separate lot of wafers, unless some IO needed to be tweaked to ensure they still operated correctly at the lower voltage. They'd be a special variant. Intel does it, and has done so for years just not to such an extreme.
It just comes back to how much money each company has to work with. AMD doesn't have money to throw around like intel does.
>>51444945
>So then why not design one?
They already have a low-power core, Jaguar. For whatever reason they haven't decided to make a 16-core or whatever Jaguar Opteron yet.
>>51444945
>Do dual am3+ or ecc?
>What the hell are they thinking?
No, they should make a dual AM3+ board that supports ECC memory, that would cut into opteron sales a bit, but those are already low as shit, and people would jump all over 16 cores for $250 with ECC support
Going a bit further than that they should also release dual/quad socket AM1 boards, for low power many core servers
http://wccftech.com/amd-ditching-bulldozer-instruction-sets-zen-high-performance-core-wip/
just found this guys
Its about changes to the instruction set for Zen.
seem like it has alot of the features that broadwell has.
odd.
I see this chip is going to be expensive as fuck.
>>51445085
That is extremely old, and its a prime example of why you shouldn't ever give wccftech your web traffic.
They claim Zen was dropping some instruction set support. The source for this claim was a forum post, a forum post where the poster in question failed to fully read the patch notes in question, then was immediately corrected.
It was all over anandtech and semiaccurate when it happened nearly a year ago.
>>51445132
So can we call this a myth and busted?
>>51436395
>pci3
>not 4
FUG
>>51445150
FMA4 FMA3 are still in Zen.
As far as I know only XOP and TBM are yet to be referenced as supported in Zen.
>>51445231
thanks anon for clearing that up.
>>51445224
That is the server variant
>>51445543
thank dog :DDD
>>51445224
Literally no one needs PCIe 4
>>51445594
Spec isn't even expected to be finalized until 2017.
>>51445594
Have we even saturated pcie 2.0 x16?
>>51445876
The theoretical limit? probably not, but we moved to gen 3 anyways, probably gives you more lanes, like with X99
