How much it is going to suck?
>>55107330
Very.
>>55107330
enough to make intel fuckbois cry in fear and desperation
It'll be decent enough competition vs Haswell-E, except in FPU heavy workloads, and it uses less power.
>>55107330
Not really much to go on, I would wait until later in the year, all you're getting now is pure speculation.
Putrid shit
>>55107330
If AMD shuts down, your Intel prices will certainly suck.
>>55107330
Very little compared to what'll happen if AMD goes under and Intel gets monopoly in the x86 market. Just look at how they started charging extra for clockable CPU's as soon as AMD left the high end market and how they bribed hardware vendors to the tune of billions to not use AMD hardware when they had the better hardware.
>>55107767
And what do you suggest? Buying shit to make sure decent product has low price?
>>55107922
>>55107330
Housefires, shit performance, pretty much what they have been doing those last years.
If it gives me the same fps as a 6600 I'll buy it.
The fact that it's taking them 5 years to release it and it's still delayed for who knows how long, tells us that it sucks.
No idea, but knowing AMD it'll just be mediocre
Yeah meanwhile Broadwell-E status: Dead on Arrival & Housefires at a price of a used car
https://semiaccurate.com/2016/06/13/intels-broadwell-e-not-released/
>>55107330
>>55107394
>>55107435
If it is any like the last stuff they released it will be total garbage despite hteir really high claims.
If they were more honest with their claims, they wouldn't have got such a shit reputation.
I think anyone can understand that they cannot seriously compete with intel who pour 10 times as much money into a new CPU generation compared to them and have their own factories.
>>55108257
Intel spends ten times as much but they get their shit out ten times quicker
>>55108314
Yes and?
Meaning they spend even more if they release new stuff more often.
Can we get rid of that nasty x86 arch/instruction set. It's 2000 fucking 16, we design something multitudes better than this.
>>55107330
>x86
>2016
You should have seen the hype on /g/ when they were developing Bulldozer. Zen will flop just like Bulldozer. It will underachieve in almost everything.
>>55107330
if it actually even gets close to Intel's new processor's in terms of speed then AMD has won, most people are still on 2500k's because Intel's new processor's aren't really very much faster and haven't been for a long time now, at this point it's not a question of whether AMD is going to blow a hole in the Jewtel's monopoly it's when
https://libreboot.org/faq/#amdpsp
>This is basically AMD's own version of the Intel Management Engine. It has all of the same basic security and freedom issues, although the implementation is wildly different.
AYYMDPOORFAGS CONFIRMED ON SUICIDE WATCH
Not a AMD hater shill thingy. But why doesn't amd just stop or whatever. Or just stick with CPU's
>>55108867
Do you understand that competition is good, right?
>>55108535
>ARM
>Risc
>ready for $CURRENT_YEAR
>>55108867
good idea everyone should just stop making technology
>>55107330
wait what, did I read correctly?
availability in 2016?
op where did you get this?
>>55108859
I have fx 8350. Am i safe?
>>55108867
>Still selling products
>Still winning large contracts
>Still has employees that honestly believe that they can make a good product
Oh yeah, they should totally give up. This is definitely not shilling btw, just an outsiders guess at what's going on inside.
>>55108976
http://wccftech.com/amd-zen-summit-ridge-launch-q4-2016/
Apparently said during their earnings call, whatever the devil that is. Personally I'm not too fussed about summit ridge, I want an AM4 APU. I want something to overclock and toy with, I don't actually need anything too powerful. Being able to upgrade to the Zen APUs after that would also be very, very nice.
>>55107330
95W in full load according to leaked info.
What is an fpu heavy workload? common shit we do on computers, not highly specialized shit please.
>>55111383
FPU: Prime 95, SuperPi, and such.
If by "FPU" you mean SIMD as is often the case today, then x264, decoding video (if you are not using hardware decoder), other multimedia software.
If such software utilizes AVX or AVX2, Intel cores will have advantage. But AVX must make a big portion of the runtime for it to be relevant. For example in x264, it is not that big factor. In x265, the AVX2 usage is better, but it is not the main performance factor, so we are talking about maybe 10% difference.
If the software just uses 128bit ops (SSE-SSE4), then Zen might not be behind at all - we shall see.
>>55108254
Jesus christ why do they fucking market this shit?
>>55108867
The GPU division is the reason they don't have to rush Zen, they been doing great
zen already taped out a few weeks ago. amd originally wanted it out in q1 2016 with zen+ coming q1 2017. take from that what you will.
btw, it takes minimum 15 months from tapeout to availability for AMD. :^)
>>55108351
>Yes and?
he's saying that when Intel spends 10 times more to release 10 times earlier, amd spends 10 times less and releases 10 years later
basically, your argument is irrelevant and if anything, it's actually against intel
>>55108145
INTEL KIKES BTFO
>>55107330
It will suck a $300 from my wallet.
>>55108363
Like what? An ARM based desktop cpu?
>>55109003
You might want to purchase housefire insurance.
Serious, not memeing
>>55113098
WTF. It doesn't even have the TDP of LGA 2011 CPUs. I wouldn't OC it because it eats a lot then, but it's no tlike it is exceptionally hot CPU.
>>55112666
Not really man. They have chips on hand so it couldn't have "taped out" few weeks ago. Tape out means you finished the design and sent data to factory (on tape medium in older days), it means start of the manufacturing preparation cycle and it takes a lot of time to get the end product.
>>55112666
They taped out in 2015. Wut you on about?
>>55113075
MIPS
POWER
>>55111514
what i mean is people talk about intel being an fpu monster and i wanted to know what i commonly use that would take a "hit" if i went amd over intel...
granted im on a phenom 2 955, im fucking sure amd or intel it would be faster no matter what, i just want to try and wrap my head around what i may miss out on on zen if that comes out for the right price.
>>55107330
>SMT
yay another Hyper-Threading -like technology that will make people think cpu has two times more cores when in reality it can only be used for a limited sets of things - fuckin learn your low level stuff and say "NO" to marketing bullshit
>>55114205
>he thinks SMT is useless
There is literally no reason to not implement it.
>>55108535
I'll buy a arm ATX board with sata ports, dual digital output and PCI buses in a heart beat for work.
Funny how in the 90's you had more choice with stuff like SPARC and powerpc and some others
>>55113036
Die.
>>55113996
except for several AMD employees who explicitly said zen hadn't taped out in 2015
>>55113224
>They have chips on hand so it couldn't have "taped out" few weeks ago.
taped out in this case meaning that they got their first wafers back from the fab, they still have to do tons of validation and will probably need to change the masks at least 2-3 times before fixing all the bugs.
>>55114205
>anon doesn't build software daily
>>55114361
It did. Last summer. SFTU AND STOP SPREADING FUD AND LIES.
http://wccftech.com/amd-zen-k12-taped/
>>55114554
>using wccftech as a source
LOL
O
L
http://techfrag.com/2015/12/09/confirmed-amd-zen-hasnt-taped-out-yet-reveals-chief-financial-officer/
who cares? just buy it if you already shill it and be happy.
When will we start seeing ZEN cpu for normal buyers? I'm not building a new PC until I can completely avoid Sheketel.
>>55114783
q1 2017 at the earliest
>>55107330
I've been using a 2500k since they were released.
Thinking of getting Zen if it lives up to the hype.
How are AMD motherboards?
>>55114706
THE FUCKING CEO OF AMD CONFIRMED IT LAST YEAR. STFU AND KILL YOURSELF.
http://www.kitguru.net/components/graphic-cards/anton-shilov/amd-we-have-taped-out-our-first-finfet-products/
She didn't lie to investors about progress. Literally end yourself.
>>55114869
>THE FUCKING CEO OF AMD CONFIRMED IT LAST YEAR. STFU AND KILL YOURSELF.
she was talking about poolaris GPUs in that press conference you idiot, zen had not taped out at that point (and several amd employees have made public statements backing this up).
on the contrary, the idea that they taped out the first poolaris samples for validation at that time seems to be correct, mostly because their first 'public demo' was about 6 months after that.
>>55114869
I've been using AM3 Socket since fuck knows when, probably 2013 not even the Cell powering the clock has died yet
>>55114894
Meant for >>55114799
>>55114869
jim keller left amd 9 months ago in september 2015, amd then made a statement directly after he left that he chose to leave AFTER THE DESIGN PROCESS FOR ZEN HAD FINISHED.
are you now trying to claim that AMD 'taped out' zen before the design had finished? how ignorant can you be?
>>55114554
That was BS, actually tapeout was more likely later. Late 2015 or early 2016. Summer 15 is out of question.
There was just some info that AMD taped out "some 14nm chip". Could have been GPU, could ahve been semi-custom, could have been something just for testing, but no, retards at that site just had to assume it is Zen, because ad revenue. Remember that GPUs and then simple CPUs like Kabini always appear first, before complex CPUs like Kaveri. Zen couldn't have been first 14nm chip for them to tapeout.
>>55114706
This OTOH might not have been accurate either, since it was just CFO speaking and for him, the topic was really tangential and he might have misspoken.
>>55114794
No, they are actually right on time with things to hit their schedula from march 2015, which put Summit Ridge "as product" to october 2016. They are now on track for release in november or december 2016.
>>55114921
>No, they are actually right on time with things to hit their schedula from march 2015, which put Summit Ridge "as product" to october 2016. They are now on track for release in november or december 2016.
no they are not, because zen did not tape out until a few weeks ago. it takes at least 12-15 months to validate and produce enough CPUs for the market. there has literally never been a CPU that went from design to production faster than that, and AMD certainly isn't going to break any records with a team of 1000 H1B indians.
>>55108254
m8 that's fucking savage
>>55107330
I'm hoping it doesn't suck, I'll be building a new computer at some point and I'm thinking about Zen since I wouldn't notice a marginal performance difference. It's going to be a lot faster than my current ancient CPU anyway which is still using AM3 socket.
I think I have no reason to choose Intel if they're both about the same and AMD is keeping competition alive. We have already seen how lazy Intel gets with their performance upgrades if the competition isn't on par with them.
>>55115017
Zen should be very competitively priced which is why it could get interesting. Even if it just means Intel being forced to lowering prices would be a good fucking start.
>>55115075
>Zen should be very competitively priced which is why it could get interesting.
zen isn't competing with intel's -dt lineups. it's going to be competing with broadwell-e and skylake-e. expect at least $500 for an 8 core chip unless AMD makes a serious design blunder.
>>55115108
The cheapest 8 core Intel starts at like £800 here. AMD won't be that expensive here.
>>55114949
bating or WTH? The showed how they have those ES in labs, some partners will get them by the end of june. The schedule said: ES available in april. They had them running on jun 1st. So there is at most a month of delay, if there even is any.
But I suspect you are just trolling and I waste my breath by not just shitposting like everybody else here.
>people regurgitating the "it hasn't taped out" meme
This all stemmed from a transcription error done by SeekingAlpha. The man speaking was an Indian(go figure) with a very thick accent.
The first Summit Ridge ES that was spotted was back in Feb. That means that GloFo already had all masks, and had run test silicon 8 months prior. Chips do not get made over night, it takes a very long time. AMD didn't just materialize Summit Ridge out of their asses and get working samples on hand.
what does tape out mean
>>55115222
>bating or WTH? The showed how they have those ES in labs, some partners will get them by the end of june. The schedule said: ES available in april. They had them running on jun 1st. So there is at most a month of delay, if there even is any.
those are validation samples from their first tapeout. it takes 2-3 respins after that, not counting the final respin for mass production, before the CPU is ready.
each respin takes ~3-4 months each time, which is why Q4 2016 is a completely inaccurate estimate for when zen will be available.
here's a talk by an AMD engineer about the cpu design process: https://www.youtube.com/watch?v=eDmv0sDB1Ak
>>55115266
Its when the masks are all finished and sent for production.
>>55115274
>I'm pulling things out of my ass: The Post.
Cute, kid.
Its been stated numerous times that 2017 would be the first full fiscal year for Zen based parts. Know what that means, junior? It means they'll be available first in 2016. Su even restated this point yet again during Computex, doubling down in saying that everything was right on track.
Who is a more reputable source, the CEO of a company. or a tech illiterate shitposting faggot kid?
>>55115323
>>I'm pulling things out of my ass: The Post.
you clearly didn't read the post before replying, everything i'm saying was sourced directly from an employee of AMD.
also
>thinking a CEO, especially an old woman, would be technologically competent enough to know if zen would actually be ready by then
this is AMD hype marketing at work, nothing more, nothing less. AMD is smart enough to know not to push out a CPU early after the phenom 1 fiasco.
>>55108254
Wait, what? Are you saying Broadwell-E was a failure based on some random overclocking tests? At BIOS default settings it ran 56 C on that same test...
>>55113191
I'm using the stock shitty heat sink that it came boxed with. Sure it's loud like a jet engine, but it never goes over 45c when gaming.
>>55115365
>Even more ass pulls when backed into a corner: The Post.
Not a single thing you said is sourced by anyone aside from your own ass. You're regurgitating the same clickbait headline originating from SeekingAlpha that everyone in the retard echochamber picked up on.
The first Summit Ridge ES chips were around at the beginning of this year. A week ago there was a B stepping chip that leaked online. Guess what? Thats final silicon.
Lisa Su has such an extensive background in IC design that the rest of your tech illiterate shitpost isn't worth addressing.
Keep spreading that FUD, kid.
>>55115412
>Not a single thing you said is sourced by anyone aside from your own ass. You're regurgitating the same clickbait headline originating from SeekingAlpha that everyone in the retard echochamber picked up on.
you clearly have trouble reading, let me link it again: https://www.youtube.com/watch?v=eDmv0sDB1Ak
>The first Summit Ridge ES chips were around at the beginning of this year. A week ago there was a B stepping chip that leaked online. Guess what? Thats final silicon.
[citation needed], the only summit ridge ES we've seen was a PCB attached to heatspreader that lisa su showed in a closed presentation a few days ago. no evidence of any functioning zen dies yet.
>Lisa Su has such an extensive background in IC design that the rest of your tech illiterate shitpost isn't worth addressing.
if she was successful she wouldn't be running a company like AMD.
>>55108145
>hurr durr da joos r bad so don bye joo stuff
Spotted the retard. Even if you're racist or anti-semitic this illogical bullshit should embarrass you.
>>55115459
hello JIDF
>>55111514
We're lucky for most things to utilize any codeset beyond SSE2. AVX workloads are a meme and nobody uses 256bit FP capability in gaymes.
>>55115518
>AVX workloads are a meme and nobody uses 256bit FP capability in gaymes.
AVX is a meme because AMD didn't make a properly implementation in bulldozer/steamroller, and intel's causes the chip to get really hot in newer iterations.
>>55115527
>causes the chip to get really hot
citation please.
>>55115274
>those are validation samples from their first tapeout. it takes 2-3 respins after that, not counting the final respin for mass production, before the CPU is ready.
That was before chicken bits. Nowadays silicon more often than not ships as A1 revision. If the original plan said ES in april and product in october, and the reality was that they have ES by end of may, then they are on track, or at most a month behind. Now if something breaks NOW, the plan might slip. But during Computes, it was on track.
>>55115684
https://hardforum.com/threads/adaptive-vcore-and-avx-temperatures-a-revelation.1777509/
normally stable haswell CPUs are always unstable when using avx2 instructions without a big increase in voltage
>>55115781
Did they fix that in Skylake?
>>55115781
Sounds like something only overclockers would ever need to worry about.
>>55115527
>>55115684
>>55115781
>>55115870
>>55115941
There is no "fix" for the issue of OC AVX stability, and lower clocks in Broadwell-E when crunching large vectors. The data paths are so large that they take a significant amount of power to switch on. Intel's solution as mentioned above is that in newer chips they utilize separate clocks for 128bit AVX and larger ops.
That many transistors switching all together in a very dense circuit is no joke. Comes with the territory when you have tons of cores though. Its just that consumer machines aren't equipped like enterprise racks with climate control.
>>55107330
I'm hoping it does well enough to prove to be a viable upgrade over my SB-E Xeon. An overclockable 8c/16t CPU w/ at least Haswell IPC for around $400 or less would be a pretty solid buy for me.
AMD has me sold on the RX480 though. Near GTX 980/R9 Nano performance for $200? I'll take one to replace my very aged GTX 660. If the RX 460 proves to be cheap enough, I might snag one of those as a slave card as well.
>>55118017
>around $400 or less
Nah, it's probably going to be at least $500, maybe even $600. It's a full 8 core chip with SMT. Intel is selling those at $1000, AMD won't price them so low.
>>55118098
500 would still be OK depending on what the platform itself offers, but 600 would definitely be on the high side, and I would only expect 600 for an 8c Zen if it proves to have ~= Broadwell performance. Keep in mind if this proves competitive it will force intel to drop the prices of pretty much all of their processors, and maybe force them to drop the dual-core dies entirely.
>>55118184
I'm just being realistic. There is hype since I actually need more cores on the system, but I'd rather have 8 of them instead of 6. I can't take their claims to heart, but around Haswell performance would be fantastic. Broadwell-E is shit and I'm not paying $1000 for a 5960x. Intel needs a wakeup call.
>>55114205
Are you a fucking idiot? Hyperthreading is useful because it turns out that the "limited sets of things" is exactly what consumer CPUs do in 99% of consumer workloads where performance matters.
>>55118229
I'd bet Vegas money on top end Summit Ridge selling for $500, with the top end Raven Ridge APU selling around $250~
Intel's lowest binned Broadwell-E is $434, the next higher binning is $617. Though Summit Ridge absolutely will not compete on IPC, it will offer two more physical cores, have a lower TDP, and is likely to have some better tertiary feature support as a value added proposition.
Intel is going to retain their FPU dominance, and even int perf AMD won't entirely meet intel's latest arch, but they might exceed Haswell in certain ops while having overall parity. That alone will make it an attractive platform.
Keep in mind that AMD explicitly stated that they know consumer desktops are not a growth market for them, Zen's bread and butter is in enterprise. 8 channel Opterons are going to be impressive.
>>55113075
https://en.wikipedia.org/wiki/Ternary_computer
>>55107330
>"ZEN"
>>55118229
Well, we'll just have to wait and see. If Zen proves to be a flop (which i'm pretty sure everyone who isnt a rabid fucking fanboy hopes it isnt), then there's always the backup option of getting secondhand Xeons off ebay. E5-2690v3 (12c) can be had for under $400, and they clock high enough to be useable as both a work chip and a daily driver.
Thats another thing I'm hoping Zen does: Forces intel to drop the prices of their goddamn Xeons, although that one is a much longer shot at happening.
I am however quite interested to see what AMD's server chips are going to do, especially the rather persistent rumor of a 32C behemoth. If it proves true Intel literally will not have anything to answer it with, as their highest core count single die is the 24C broadwell-EX die, and AMD has had the capacity for a very long time to do 4P and 8P systems.
>>55118319
AMD's biggest selling point for enterprise is something that you probably wont see in any benches, unless Anandtech buys a 4 socket systems and does a super indepth analysis.
The Freedom Fabric IP they acquired is basically a super low latency PCI mesh that scales with multiple sockets better than anything else on the market. Thats what they originally planned for Skybridge, and thats why they had so much faith in ARM for the future data center.
A Zen based Opteron might offer 80% of the performance of a Broadwell Zeon, but socket to socket scaling it surpasses is, so eventually a rack full of Zen chips comes out on top. There is a ton of performance lost in signaling with long wire lengths.
>>55118359
Pretty sure anand has at one point or another done 4P reviews. I know a bunch of other sites have done it.
IMO AMD will actually have something strong for the data center, because not only is Zen expected to scale up really really well, but AMD also has the extremely strong power control IP they developed to get a handle on the Construction core's heat and power consumption. The very same power control IP that, alongside some architecture tweaks, allowed the conversion of the power hungry hot running hog that was Bulldozer into the power sipping Excavator, which is a pretty massive improvement.
As we all know, if theres one thing data centers like more than anything else, its increased power efficiency, and should AMD implement their power control IP into Zen (I dont see why not, Resonant clock mesh alone got something like a 15% reduction in power per clock while simultaneously allowing for increased clock rates), they'll definitely get data centers to give them some serious thought
>>55118582
All of that PowerTune IP is what let the Fury X have a 220w~ average power draw. Under heavy compute loads it could jump up to 350w, which is exactly what you'd expect from looking at Hawaii. For being 28nm silicon still that is a massive improvement in energy per op. They got progressively better and better at gating, fine voltage control, and a host of other things.
All of that is going to be baked into all of their future parts, and they likely have things a few generations newer than what we've already seen.
Its still kind of impressive that the PS4/Xbone SoCs are both Sea Island GPUs with PowerTune IP that didn't reach market for consumer cards until Fury. Though looking at their media playback power draw its obvious they still had some bugs to work out. Still, those chips were originally designed back in 2011 which means they have major advances ready to field if they have the money. Those SoCs are like 100w and 65w respectively. For 2011 designed 28nm planar bulk thats mind blowning. The PS4 has a full blown Radeon 7870 on die with only 2CU binned off. There are 8 Jaguar cores sharing that power/thermal envelope as well. They're clearly on to something over at AMD's labs.
>>55118641
Yep, and thats why I'm optimistic that AMD will get marketshare back in the enterprise area. Broadwell-E was a bit of a flop in terms of power consumption and heat, and with the barely there increase in performance so typical of Intel these days (I personally maintain that its because intel refuses to re-design their core from the bottom up and instead insist on continuing to slap shit on a heavily modified P6), data centers are probably looking into grabbing some Zen parts.
The big question in all of this though is how Zen clocks. Having a high IPC design (and lets face it, pic related is a high IPC design) doesnt mean shit if it wont clock well, and thats one of the things AMD has been really tight-lipped about
I feel like it will just be more of the same from AMD. I'm still using a 2500K and it hasn't yet let me down, so I might end up just waiting for Skylake-E if Zen isn't amazing.
