Zen
Images are sometimes not shown due to bandwidth/network limitations. Refreshing the page usually helps.
You are currently reading a thread in /g/ - Technology
You are currently reading a thread in /g/ - Technology
Who else hyped about Zen architecture?
They don't have to beat Intel on IPC if they can get close and offer quad cores with hyper threading at a lower pricepoint than intel does.
This guy made a good analogy of how Zen could be succesful even if it doesn't beat Intel on IPC
https://www.youtube.com/watch?v=OvF3BJTLgRQ
I am still on a 2500k right now and I'm kind of sick of intel milking the market with marginal 5% performance gains each year while fucking over people who managed to overclock non K processors.
If Zen delivers on its promises I would definately be in the market for a good quad core with hyperthreading
>>53341359
If they offer a quad core with ht at i5 prices the i7 line is finished.
But I'm expecting 2 products around i5 price range. Quad cores and tri cores with ht.
>>53341695
If by ht you mean hyperthreading AMD cannot do that since Hyperthreading is licensed by Intel.
>>53341359
>totally new
Dude like that's some totally radical shit dude
Their marketing department needs to be shot
>>53341751
It is a new architecture you retard
>>53341737
Not true.
AMD is using smt for ZEN instead of cmt. Be prepared for ZEN processors with hyperthreading anon!
>>53341737
This
>>53341768
They have basically the same thing with a different name
>>53341695
Isn't the lower end I7 line basicaly quad cores with HT too?
>>53341774
I admit my ignorance of that new name since its been a while since I last looked up zen.
What name are they going with?
I would love it if they offered a 6 core 18 thread processor, but I know I'll never be so lucky ;_;
Since Intel is hyperthreading AMD should use megathreading or something better than hyper anyways.
I'm still on an i7 920.
I desperately want Zen but I'm not sure I can wait another 6-8 months.
What's the i7 920 equivalent on Intel's current line? I want it to last another 7 years.
>>53341788
Exactly.
Why buy an i7 when zen is almost the same but almost half the price?
I realise fx8350 and i7 3770 was similar, but back then a lot of people chose the i7 because of efficiency.
>>53341791
>I would love it if they offered a 6 core 18 thread processor,
>>53341804
3 threads per core, madness
>>53341798
Let's wait how Zen actually turns out before speculating on what to do if they are the same.
I am cautiously optimistic. There does seem to be a lot of evidence that zen will be competitive but I remember all the hype and shilling behind bulldozer. People want to believe.
So is AMD done if Zen flops?
>>53341810
I've seen 8 core powers with 8 threads per core, why can't Intel or AND offer more than one thread?
>>53341835
Bulldozer was very competitive in tasks that took advantage of as many cores are available. Like rendering.
>>53341838
No.
>>53341850
Intel or AMD*
>>53341835
>>53341838
As much as I want to see AMD succeed the hype was the same for the Fury cards and they didn't beat Nvidia or offer lower price. So I don't hold any expectations. Tho is there any word on whether there will be FM2+ ZEN CPU's? Might buy a replacement for my X4 860K.
>>53341359
cool video op.
i'll subscribe to your patreon.
>>53341850
What would be the point? It's not like threads can't be done in software, you don't really need special hardware support. 8 threads are still overkill for nearly all applications, so why try to make even more of them?
>>53341835
The thing that makes me optimistic is that they don't have to actualy beat Intel to make a huge impact.
Intel has shot themselves in the foot by milking the market dry with their incremental performance upgrades and high prices (which I don't blame them, because AMD's CPU's were so shit there simply was not competition).
Now there is a huge opportunity for AMD to hurt them with stuff like DX12 more cores will actualy start to make more sense.
Everything looks good, Jim Keller, 14nm licensed process from Samsung, 40% extra IPC, hyperthreading finaly.
I for one can't wait and will happily support them if they bring out something good because I'm starting to detest Intels business practices more and more. I was forced to stick with them until now however because the competition was so horrible.
>>53341877
Yeah the guy is really good, look up his video on Polaris vs Pascal, it's really good too
>>53341891
>stuff like DX12 more cores will actualy start to make more sense.
Wasn't DX12 supposed to make CPU-power less important?
>>53341870
Fury x does offer Titan x performance at half the price though.
And it's a pretty big upgrade from a 290x.
How exactly are the fury cards a failure?
>>53341931
Yes.
And it also scales up to 6 cores very well.
That means I can hold on to my 8320 for a few more years.
>>53341931
No, it makes spreading the load across more cores easier and more effective.
So it makes single core performance less important, in which Intel is still the best
2 core Zen = Celeron tier cost but i3 tier speed
4 core Zen = i3 tier cost but i7 tier speed
6 core Zen = i5 tier cost but Xeon E5 tier speed
8 core Zen = i7 tier cost but faster than anything Intel
>>53341359
i am
not hyped but hopeful they still make to deliver good enough improvement (and a good cpu overall) to keep them afloat
>>53341796
i7 5820K
>>53342020
>Xeon E5 tier
What is this tier supposed to be?
There is a ton of very different cpus falling under that umbrella
>>53341796
Why upgrade now?
I'm still on my i7 870 and it's running perfectly fine.
In fact, overclocked it's about as fast as an i5 4460, if not faster. 4.2ghz with hyperthreading on.
You'll be fine for a few more months anon.
>>53342071
>current
>Broadwell
Why?
>>53342071
400€ :(
>>53342147
not him but
>no bending
>14nm
>gud igpu
>high hipster factor since 5xxx is somewhat not popular
>intels starting with odd number are better
>>53341936
>Titan X
Ayy lmao nobody cares about the Titan, and the cheapest GTX 980Ti is exactly the same price as the cheapest Fury X (639€) and most benchmarks I have seen show the GTX 980Ti as the better card. I guess we will have to wait for the DX12 games to know for sure.
>>53342138
I can't overclock for shit on my current mobo. 3.3Ghz is as high as I can go.
I'd like to upgrade my whole system apart from my 7970 and get a decent 20-30 fps boost until Polaris.
The next 8 months until Zen comes out is the time I can invest and game.
I'm torn.
>>53341936
Its b8
>>53342235
Completely missing the point tard. There was no 980ti until Fury X was announced
>>53341767
found a shill
>>53342376
Found an ignorant dumbass
>>53342079
He probably means 6core hedt tier, same for 8core and intels 8cores going against each other.
My feeling in the end though is similar to how it is now, just not as severe. Intel winning single core and fp ops whilst amd offers more value in the multithreaded side of most things.
>>53342020
thing is, intel has had a fat margin on their chips for a really long time. They probably can cut their margins on older chips to compete with amd IF amd finally gets something worthwhile out.
>>53342597
to continue, ZEN will either float or sink the amd's cpu market. If it flops, it's kinda hard to see it recover anymore.
Hopefully it's going to be good.
>>53341359
Waiting, watching, and wondering. Someone please explain to me with all this hype, all the built up demand why Zen won't be as expensive as Intel. There is plenty of reason to believe this will be a paper launch, with insane markups. Skylake is not exciting, but Intel can't fab enough to keep up with demand. Zen is hyped to death already, a full year ahead of release.
I don't give a shit about offering cheaper quad cores. If they can offer cheaper hexacores or octocores at a reasonable price without the IPC being two generations behind in performance, I'll give Zen a serious consideration over Skylake-E.
>>53342701
What would a target msrp be for you?
6 core $380 8 core $ 600 ?
Why can't AMD just start manufacturing ultra-low power AMD64 SOCs for mobile? It's not like the talent isn't out there.
It nearly feels like management is being paid heavily by Intel just so they can get a monopoly on the x86 market.
I haven't been excited about anything AMD has done since around 2001 or so.
There has been no reason to buy their shit since then unless you were on a tight budget.
>>53342760
Less than 600 for an octocore would be pretty great.
>>53342618
>Someone please explain to me with all this hype, all the built up demand why Zen won't be as expensive as Intel
Because bulldozer is such a huge fucking chip, those piece of dogshit cpu's are way more expensive to produce for AMD than intels cpu's are. and even despite of that, they are selling them for bargain prices because they are so bad compared to the competition.
Because Zen will be built on 14nm they will be able to produce much more dies on one wafer. meaning the production costs will be dramatically lower.
>much lower production cost
>much better performance
Basically AMD is in a perfect position to position these cpu's at an attractive price point and still make a shitton of money.
>>53341359
>They don't have to beat Intel on IPC if they can get close and offer quad cores with hyper threading at a lower pricepoint than intel does.
That would only make them attractive as a budget option. If they can get their IPC "close enough" to Intel they could provide attractive options by providing more cores at the same price point, as in sell a 6 or 8 core CPU with SMT ("HyperThreading") for the same or lower price of Intel's unlocked mainstream i7.
Why would I upgrade from my 4.8GHz+ 4790K to Zen if the only thing it brings is lower price? They need to have some sort of performance edge as well if they want it to have a large impact.
Also HyperThreading is just Intel's name for their simultaneous multithreading implementation. There are other CPUs that use SMT too, just not in the consumer x86 space. Zen will have SMT too, though it won't be called "HyperThreading" since that's Intel's name.
>>53341359
Unless they beat Intel at efficiency, I couldn't care less.
>>53341850
>Bulldozer was very competitive in tasks that took advantage of as many cores are available. Like rendering.
But most home users didn't care too much about that sort of thing. Like it or not, the main driver for high performance on consumer desktop PCs are video games and at the time Bulldozer came out it was inadequate for the task compared to Intel.
>>53343879
>If they can get their IPC "close enough" to Intel they could provide attractive options by providing more cores at the same price point
Which was exactly my point.
They might not pursuade people who already have an I7, but they will be the more attractive option for people still on an I5 or looking for a new CPU in their build.
Maybe their top of the line 8 core processors with HT will even challenge the top I7's.
They don't ship them with integrated graphics so that could mean they would be a big chunk cheaper than their Intel counterparts.
They won't beat Intel on IPC with a new architecture while Intel has perfected its for so many years. But they can get close
>>53343880
Are you some kind of enviromental activist?
>>53341931
It reduces overhead, so you could say that. But that's just overhead relating to sending work to the GPU, a game can still be CPU intensive. For instance, Ashes of the Singularity, have a look. Running at 4K with 2 290Xs, it's CPU limited on a 4770K.
While running the benchmark I saw it hit 100% CPU load very frequently and even when not at 100% it was very close.
>>53343919
I simply want the best system and efficiency is a good indicator for it. Unless you're a power user (these don't use consumer cpus) Even some CPU from two years ago is good enough, so the only reasonable excuse to upgrade is being to able to do the same/bit more while using much lesser power.
>>53343907
I'm looking for an upgrade from a 4790K to something from Broadwell-E or Zen, I definitely hope AMD can be competitive.
I use Steam In-Home streaming pretty often and the CPU overhead is significant, even with hardware encoding. But it sure as fuck would be amazing to be able to run software encoding without a large hit, since it provides much better quality.
>>53343961
>I simply want the best system and efficiency is a good indicator for it
That depends entirely how you define 'best', it sounds like you define it as 'Most power per watt' in that case, right now you are right.
>Due to Zen integrating previously external motherboard components in an SoC-like nature, Zen's AM4 socket will unify the future variants of older AM1, FM2+, and AM3+ sockets.
Agh what does this mean?!
I'm looking forward to its release. I'm hoping it's a success as it will keep prices down for everyone.
>>53341931
Will AMD's moar cores strategy finally pay off? After years and years of being mercilessly mocked, will they get the last laugh?
>>53344020
The motherboard becomes slightly less complicated because the CPU includes bits in there that previously weren't. I don't know what those bits are, but hopefully that helps your low-literacy retard brain understand.
I'm only interested in IPC, since it's basically the most important thing nowadays.
So unless Zen defeats Kaby Lake when it launches, it is going to be dogshit to me.
And we all know it won't reach Kaby Lake levels, hell, I even have my doubts it will reach Haswell.
AMD just cant win. They were too much time behind. They cant keep up right now.
>>53341835
It was horrifying. I was going to make a new PC from scratch at the time and decided to wait for the Dozer. It was just embarrassing having their Phenoms beating their new CPUs. That shit could really be their undoing, they might just die if Zen isn't a killer. Just beat my Ivy Bridge in single threaded applications and I'll gladly get an 8 core.
>>53344103
It seems like alot of their earlier decisions are starting to pay off in future tech.
Asynchronous shaders in their graphics card, more cores in their cpu's.
Although I don't think the more cores thing was a deliberate choice aimed at future software developments. More of a necessity because they were so far behind in IPC compared to intel
>>53344184
>dogshit to me
nobody gives a fuck about a pedophile's opinion
>>53344184
>I'm only interested in IPC, since it's basically the most important thing nowadays.
It really isn't though, plenty of things are capable of using multiple cores nowadays, and that's only going to increase.
>>53344020
It means it'll be more like Intel motherboards.
I am still on a Q9550 and I am suffering. I will wait for AMD but if they fuck up again then I'll just get a Skylake i5.
>>53341359
Summit Ridge chips aren't going to be cheap, Raven Ridge will be priced to compete against intel's mainstream chips.
>>53341737
Hyperthreading is a pointless marketing name for intel's implementation of SMT.
>>53341850
SMT offers diminishing returns to a crazy degree. Every additional thread offers less performance uplift than the last, and when you're sacrificing die area and power consumption to achieve this you reach a point where its pure insanity. Look at Anandtech's benches comparing a POWER8 rack vs a rack of Xeons. 36 Xeon threads vs 100 POWER8 threads with superior clocks. Additional hardware delivers consistent large performance gains in heavily threaded workloads, virtual threads do not. Even in some of the most intensive synthetics intel only sees a 30% improvement in throughput per core from SMT. They understand that the approach only makes sense if you spend as little die area on it as possible.
>>53344020
AMD's upcoming Summit Ridge CPU, and Raven Ridge APU all use the same socket, AM4.
Excavator based Bristol Ridge and Stoney Ridge also use AM4.
>>53344184
There is literally zero chance of Zen outperforming Kaby Lake, or Skylake, or even Haswell across the board.
Lisa Su directly said that their Zen Opertons are going to be competitive in 80% of server workloads, and that means they're going to lose out in the other 20%.
>>53342225
>>53342147
5820k is haswell
>>53341695
>>53341737
>>53341768
>>53341774
You people mean multithreading, not hyperthreading. Exactly how AMD will implement is not known AFAIK, but it probably won't be so called clustered multithreading (a "module" or "cluster" with two integer cores and one floating point core) like Piledriver
>>53341695
I know the 6300 is basiclly tri module, but I hope not. I guess that I was expecting four core as a minimum, but now I'm thinking that's optimistic.
>>53341359
That guy's accent is fucking atroious.
>>53346870
They mean SMT, exactly was was explicitly stated.
SMT only works one way. Its even one of the first bullet points in the slide the OP posted from AMD's FAD last year.
>>53341359
Still waiting for either Zen or a good ARM64 board for my next desktop. Basically anything that will let me avoid buying intel for a few more years.
>Release date is October
So that pretty much means you'd have to wait until 2017 for decent availability and prices. GJ AMD.
>>53341359
Cautiously Hyped. Even if Intel pulls a hat-trick at the last second (massive price drop, crazy new IPC/clock speed) AMD has the potential to finally light a fire under their ass
>>53346905
Hyperthread is a type of multithreading; not all types of multithreading are hyperthreading.
Hyperthreading is proprietary Intel technology. AMD will not use hyperthreading.
>>53347665
>I'm a tech illiterate shit eating retard
Hyperthreading is nothing but a marketing name for SMT.
Intel did not invent SMT.
Zen is an SMT arch.
IBM's POWER is an SMT arch.
The latest iterations of MIPS are SMT as well.
Go back to /v/ and stay there forever, you dumb little faggot.
>>53347707
>Hyper-threading (officially called Hyper-Threading Technology or HT Technology, and abbreviated as HTT or HT) is Intel's proprietary simultaneous multithreading (SMT)
https://en.wikipedia.org/wiki/Hyper-threading
tl;dr You're a moron
>>53341850
>I've seen 8 core powers with 8 threads per core
Because POWER is REALLY fucking expensive industrial-grade hardware.
Because 8 threads per core is totally useless for normal users.
Probably cool for servers, but I don't need high performance.
>>53347734
>literally proving yourself wrong with a wikipedia article
Hyperthreading is a marketing name. *ALL* SMT works in the same way, ops are tagged in registers, and get executed on nonbusy ALU. There is no fundamental difference in this from ImaginationTech's Warrior cores to intel's Skylake. It is the exact same process.
Zen is an SMT arch, exactly as has been said numerous times now. If you weren't such a blind tech illiterate retard you'd see that AMD themselves made this perfectly clear.
>>53347748
Power8 supports up to 12 threads per core in the highest binnings, and hits 5ghz.
IBM is losing it.
>>53347074
Clock speeds aren't going anywhere until radically new materials are used.
The only significant gains to IPC Intel has made are with vector/parallel/floating point architecture - everything else is a simple optimization over last year's.
x86 is pretty much tapped out for them and we'll be lucky if a general 3-5% improvement every release is even sustainable.
Zen = similar IPC and overclocking/efficiency to Intel
But better prices and more cores
If you have a 5820k don't upgrade. Only upgrade if you currently have a quad core.
I'M FUCKING READY
>>53348369
>tfw I might finaly see a decently priced quad core with SMT to replace my 2500k from a company I actualy want to support with my money
>>53342834amd sold their mobile market to qualcomm with a non compete in mobile
>>53341881
It would be incredibly specific use case but lots of HW threads would be great for environments with high latency IO
