ZEN will save AMD they said

>ZEN will save AMD they said
>ZEN will beat skylake they said
>ZEN will bring back the competition they said
>ZEN will [insert any promise made by amd fanboy] they said
i dont want to be on an amd hate train but COME ON

Full article:
techspot.com/news/65945-leaked-benchmarks-amd-zen-falls-short-intel-haswell.html

Other urls found in this thread:

cpu.userbenchmark.com/Compare/Intel-Core-i7-4790-vs-Intel-Core-i7-6700/2293vs3515
techspot.com/review/1173-doom-benchmarks/page5.html
guru3d.com/articles-pages/total-war-warhammer-directx-12-pc-graphics-performance-benchmark-review,8.html
guru3d.com/articles_pages/far_cry_primal_pc_graphics_performance_benchmark_review,9.html
guru3d.com/articles_pages/anno_2205_pc_graphics_performance_benchmark_review,9.html
guru3d.com/articles_pages/rise_of_the_tomb_raider_pc_graphics_performance_benchmark_review,9.html
i.imgur.com/5JEenmU.png
i.imgur.com/KVnTvmS.png
anandtech.com/bench/product/1684?vs=288
anandtech.com/bench/product/1684?vs=186
twitter.com/NSFWRedditImage

This isn't a fair benchmark, a fair benchmark is SuperPi and x87 code.

Btw it was proved that isn't actually ZEN, it's some Intel chip with a modified CPUID, the real ZEN is in fact much slower

DELET

I don't even know who's shitposting who ITT

...

look at the clock speed, and this is not the final product (minor tweaks inbound)

still, skylake beats the fuck out of 4790 (especially in gaming), so ZEN, even after tweaks and shit will have a hard time catching up.

didn't devils canyon i7 have better performance in some games than skylake i7?

>trusting leaked benchmarks
>ever
As I recall, leaked benchmarks also "proved" that bulldozer was better than sandy bridge.

Next step in the AMD Defence Force Playbook:

>ZEN wasn't even meant to compete with Skylake, it's a Haswell rival
>We never said it's gonna compete with Skylake, that was Intel shills trying to make us look bad
>It's gonna be cheaper, etc.

Just like their backpedalling after the 480 release

Next you'll tell me Skylake has more than 5% higher serial performance than Haswell so it matters that AMD is "2 gens behind"

...

+1 If you want AMD to die alrdeay so we can get actual competition

Clock by clock comparison AMD should actually be stronger. That's if Zen can overclock but AMD CPUs besides being hot never had problems with that.

Anyone find it extremely strange that AMD keeps using this game for benchmarks when it only has a few thousand players?

For only $150 we're offering a pretty fair deal

>skylake beats the fuck out of 4790 (especially in gaming)

Any examples outside of RAM bandwidth hungry games? Even Sandy Bridge-E will smash a 4790k in Fallout 4 just because of the quad channels.

Guess what, zen will have ddr4 too.

like what?
VIA?

>ZEN will beat skylake they said
I remember. They said that Zen will be 18 cores, use 5W (total, not per core), and operate at 65GHz. Anything less than this, and Zen is a total and absolute failure.

And it must cost $40, too.

If you read the leaks carefully, the AMD chip is 1GHz slower than the Intel ones.
And these are engineering samples- engineering samples typically run slower than the release, simply because it's just a 'rough cut' of the processor.

fuck off tripfaggot

>Between haswell i5's and i7's

Considering I still struggle to find a 4790 below 300€, if amd prices this one at 250/300€ they might have a pretty good deal here.

Emphasis on "if".

if the cost matters, you're poor

or, intelligent buyer.
one of the two.

>not being born into millions
what a fucking loser lmao

VIA is almost as fast as AMD

>You know you don't become a richfag by overspending on shit you don't need, right?

that cherry picked chart puts Intel at the bottom though

>euro
lol good luck
Shit that runs $400 here is like 800 euros in cuckland.

Joke's on you, I'm on a Haswell i7

If i can't go zen with an itx board I'll go for Intel to be honest

gonna need some sauce for that

why isnt ARM going for high end desktop cpu's? They seem to be getting real good at mobile dept, whats stopping them from building proper desktop chips?

Insignificant market compared to servers and mobile.

Because making lots of simple cores instead of few powerful ones is a good strategy for mobile, not so much for servers. And it absolutely sucks ass for desktops, as Bulldozer demonstrated.

>Engineering sample
>Unknown clock rates
>Unknown final clock rates
For all we know they could have run that test with half the cores disabled

AYYMD IS FINISHED AND BANKRUPT

>ZEN wasn't even meant to compete with Skylake, it's a Haswell rival
This was always the point.
>We never said it's gonna compete with Skylake, that was Intel shills trying to make us look bad
Again, this was always true, no reliable source ever said it was meant to compete with Skylake.
>It's gonna be cheaper, etc.
And this was also always true.

You basically took a bunch of valid points (points 1 and 2 are the same thing rephrased) and thought that maybe if you added meme arrows it would look silly.

>Just like their backpedalling after the 480 release
What backpedalling? What did it fail to accomplish? Everything AMD promised came true.

...

every amdshill on Sup Forums ever was sperging how Zen is going to be a skylake for cheaps

except not maybe like 20% more depending on region

>every amdshill on Sup Forums
Ok, and /x/ legitimately posts about how the Earth is flat, how does this constitute a failure by AMD?

i was expecting haswell-e for $500 cutting the 5960x's price in half.

...

AMD will never go bankrupt because Intel will just keep pumping money into them so they have something that looks like competition.

I think AMD is doomed and they will always be behind Nvidia.
Sad but true

>its lies

Yes.
It's not a big deal because Intel focuses on energy saving
I hope Zen won't be 220 W like Bulldozer

That benchmark is bullshit. It either proves that Zen has 80% higher IPC than Piledriver, which would put it in the same ballpark as Skylake, or it proves the complete opposite.

This is why you don't use games to benchmark CPUs.

>ZEN will save AMD they said
Partially, along with Polaris/Vega and Xbox, Playstation, Apple deals
>ZEN will beat skylake they said
No one said this except Intel shills, shill
>ZEN will bring back the competition they said
It will. Yeah, maybe not for the highest end parts, but for the vast majority of the market, yes.

>Op's pic
As long as the final product is clocked higher, it'll be fine.

No, you fuck off, user.

That 5960 isn't at stock clocks, dummy

amdrone pls. literally half a year ago every tech site speculated on how zen could perform against skylake. Sup Forumsg was NOT an exception.

That picture doesn't mean what you think it does.

Notice how every link saying that is from a forum, reddit or in one instance wccftech. You don't have a single reputable source. AMD NEVER in any way said Zen was meant to beat Skylake. Anything that other people speculated was their own fault.

99.999999% chance these AoTS bench is legit. The guy who posted them is a Chink, and someone allegedly stole a Summit Ridge test system from AMD's Computex event in Taipei.

The performance shown isn't necessarily bad. We're looking at an engineering sample with unknown settings. AoTS does scale with RAM speed to a high degree, the CPU could have turbo disabled, NB could be running at half rate, it could be mismanaging the SMT threads. All of that and its still substantially faster than the 4ghz FX 8450 at a much lower clock.
AMD only touted a 40% uplift in IPC over Excavator, and here we're seeing something like an 80% uplift in total performance vs Vishera.

>oh no, an engineering sample that's clocked 25% lower than the competitor's chip is 12% slower than it!
>must be a terrible chip!

I'm not even exaggerating right now. This is way better than what I had hoped for. If the TDP is the same, price slightly lower, I'm in.

>a 2.8~3.2GHz Zen chip beats a 4.0~4.2GHz FX-8350 by 16FPS
>somehow this is bad

It has a 95w TDP for the 8 core part and final clocks will be higher.
Don't count on it being priced below a mainstream i7.

I checked some of those links in your screenshot.

Literally everyone is saying it will probably only compete with haswell or come close.

Thats what the general consensus was. 40% IPC over FX series. AMD screenshot showed just that, roughly 38% better than 8350. Note it was a low clock variant too. Higher clock variant would no doubt gain more in performance, in a roughly linear fashion.

lol.
Are via actually around? Fuck they were terrible.

That doesn't sound too bad tbqhwyfam. Luckyly my 3570k doesn't really need an upgrade, so I can just wait and see what happens.

>currytech """"leaks""""
>promises made by AMD

...

I checked the link, this was a 8c16t ES variant. The low clock is due to the core/thread limitation. I reckon a 4c8t variant will spout around 3.5-4Ghz base. That would be roughly 10-25% faster base. Not sure how well that translates to the ingame metrics, but I suspect it would be anywhere from 5-15% faster. If we assume 10% boost in the game benchmarks, this would put the 4c8t variant (3.5-4Ghz) close to i7 4790 tier on base.

Now its just a matter of waiting on what the overclock potential would be.

that's pretty good. if the bench equates it will be significantly beating the best most common tier of high end intel cpu for gaming eg the four core i5's such as 2500k, 3570k, 4760k, etc.. considering nobody buys i7's for gaming that is.

Does this game have trouble pulling more than 6 threads? Clock the Zen core slightly up, drop it to 4 cores and it would do even better.

>87% improvement going from 2 to 4 cores
>41% improvement going from 4 to 6 cores
>14% improvement going from 6 to 8 cores
AoS is worthless as a CPU benchmark

Engineering samples don't have final clocks. The top binned chip will probably have a 3ghz "base" clock and an all core turbo around 3.3-3.5ghz.
The design allows for very fine control over clocking and gating which everyone seems to be conveniently ignoring since its fun to shit all over AMD. Socket AM4 has tighter power delivery requirements than any socket out there, including intel's 2011v3.
Summit Ridge will have pstates set depending on how many cores are loaded, and how high their utilization is. A full CCX can be gating off in lightly threaded workloads, and clocks will be pushed higher to make full use of the 95w TDP on 4 or fewer threads. I guarantee they won't operate at their conventional base clock all the time either. They'll throttle down to almost nothing by default to bring idle power down.
Zen cores and the CCX scale, and socket AM4 demands absurdly fine grain power delivery to support it.

It can make use of more than 6 cores, but diminishing returns is always law.

Its highly sensitive to memory frequency and timings too.

>skylake beats the fuck out of 4790 (especially in gaming)
no it doesn't

it was just another 5% boost to performance as it is with every new generation from intel

hardly beating it the fuck out

There's like 3-5% of performance difference between those chips in general(no cherry picking).

ZEN only has to beat 2500k to be successful. You guys have completely forgotten how much CPUs used to cost when AMD was still in the game.

Realistically they need to beat the i7 3960X and i7 5960X in most well threaded workloads. Thats their target competition.
If they can match that then competing with mainstream i5s and i7s is only a matter of pushing clocks high enough with fewer cores. The Raven Ridge APUs is AMD's mainstream platform.

Summit Ridge = i7E
Top end Raven Ridge with HT = i7
Mid binning Raven Ridge = i5
Low tier Raven Ridge = i3
Raven Ridge with no IGP = Celeron and Pentium

Its actually not worthless, but rather normal.

Most games don't scale well past 4 cores. The fact that Ashes even scale at all even on 8 core is pretty good.

Also recheck your numbers.

>2c -> 4c
Thats 89% on game scaling and 100% on core scaling.

>4c->6c
thats 40% on game scaling and 50% on core scaling

>6c->8c
15% game scale and 33% core scaling

Its scaling appropriately. With diminishing returns hitting.

It needs to be as fast as/faster than SB, have more cores and OC well. The rest is insignificant.

I disagree. Pretty much all recent games can utilize and scale beyond 4 threads. Intel has realized that as well. In fact, there were the ones hindering the progress.

>believing in such wide benchmark variations when processors are all exactly the goddamn same these days thanks to a fucking wall they all slammed against years ago
Why are you all so gullible as to believe Intel's lies?

AoS does not scale well beyond 4 threads and does not scale at all beyond 8 threads. This makes it completely worthless to benchmark a 16 thread CPU.

>mfw AMD have been trying to beat the i5 2500k for 5 years now
LOL

MAKE WIDE CORE, FETCH MANY INSTRUCTION YA

cpu.userbenchmark.com/Compare/Intel-Core-i7-4790-vs-Intel-Core-i7-6700/2293vs3515

Roughly only 5-6% better.

There's not much difference.

Name one recent game that has good scaling past 4 core.

DOOM is supposed to be really optimized, yet even that can only scale upto 4 core

Total War Warhammer is a CPU intensive game and supposed to be well optimized too. However even that isn't scaling well.

>techspot.com/review/1173-doom-benchmarks/page5.html
>guru3d.com/articles-pages/total-war-warhammer-directx-12-pc-graphics-performance-benchmark-review,8.html
>guru3d.com/articles_pages/far_cry_primal_pc_graphics_performance_benchmark_review,9.html
>guru3d.com/articles_pages/anno_2205_pc_graphics_performance_benchmark_review,9.html
>guru3d.com/articles_pages/rise_of_the_tomb_raider_pc_graphics_performance_benchmark_review,9.html

Reality is, CPU scaling in Ashes is pretty good compared to its peers.

Most games don't really scale well because DX11 mostly hammers the main thread.
DX12 helps, but so far it hasn't really shown it can effectively use more than 6 cores.

them's leaks nobody cares

did anyone notice these two?

this one is i.imgur.com/5JEenmU.png - ES 1D2801A2M88E4 32/28 and scores 5,300.
this one is i.imgur.com/KVnTvmS.png - ES 2D2801A2M88E4 32/28 and scores 3,100.

they're both ran in standard, both at the same resolution, and both 8 core 16 threads, both 2.8 / 3.2ghz. but they're different ES samples. wccftech does touch on this but they decided to completely ignore the 2D2801A2M88E4 results. same with guru3d.

>There doesn’t seem to be any difference at first sight between the 1D and 2D variants. They are both purportedly clocked at the same rates and the only difference that we can see is the fact that the 2D variant fares much worse in terms of performance. ..................
>In any case, because of these reasons, I decided to focus on the 1D variant for our comparison.

clearly these are indeed ES samples and being ES samples we cannot truly gauge zen's performance since they have TWO ES samples floating around with drastic performance differences between the two. 2D2801A2M88E4 scores worse than amd's very own bulldozer and the other, 1D2801A2M88E4, takes 8 cores and 16 threads to slightly lead a 4 core 4 thread haswell i5. we don't know anything about the setups or the ES samples themselves. we don't know how many cores where actually activated, was amd's hyperthreading activated, the true frequency rates, what instruction sets activated, tdp levels allowed, or anything of the like. With two ES samples that appear to be similar spec tested with such a wide gap of performance differences between the two, we can only take these benchmarks with a grain of salt.

this. I paid like $200 CAD for my 2500k. now that's the price of an i3. buying a CPU now gives me as much bang for my buck as it did in 2011 and that's fucking retarded.

I know the currency took a hit but seriously.

Engineering samples are test beds. They will try out different configs and try to find out which ones work or not. The lesser performing will be dumped infavor of higher performing design.

The point being that games are not a valid benchmark for 16 thread CPUs

>480 comes out
>doesn't beat GPUs at twice the price
>b-b-but leaks said it would beat the 980 that's false advertising isn't it?

you fragile little shit

They have to do a lot more than that. If they offer Zen for about $200, with this current market of Intel CPUs not falling in value, it's a failure.

I got my 6600k for $200.

thanks for restating everything i wrote.

ES samples are ES samples.

The slower one might have bugged microcode, low frequency NB, disabled cache, missing instruction sets.
They're ES, they have bugs.

But look how close they got:
anandtech.com/bench/product/1684?vs=288
anandtech.com/bench/product/1684?vs=186

(Carrizo)Excavator with no GeAPM, no throttling, no L3, and only half the L2 present in Steamroller. Its ends up slightly faster than the Phenom IIs in some things. Sandy Bridge is 20%-50% faster excluding heavy FPU bound workloads.
The performance differential would be slightly smaller if the APUs had an L3. So it looks like it took AMD about 4-5 solid years to finally surpass their own prior arch, which is still that far behind Sandy Bridge. A tremendous issue holding back BD performance was always cache throughput and latency.

Finally surpassing Sandy will be a big milestone for them.

Chinks aren't a valid source. Wait for the final product before you start shitting everywhere

thanks for restating everything i stated.

ES samples are ES samples and take them for a grains worth of salt.

The same CPU outside US and A is most likely 50-100% more expensive though. AMD might win everywhere else then.

AMD is trying to market the cpus for gamers. Gamers are the most vocal and enthusiastic about new cpus.

But you're right. 8c16t game benchmarks don't work well because they dont scale well in games, they also hold back the clocks because of the core/thread counts.

No, you said there's no conclusion to be made. I said the lower ES will be dumped in favor of higher ES. So where numbers count, the higher one would be a more valid approach to proper benchmark evaluation rather than saying "lul theres two and we cant make sense of anything"

>editorials
>speculation
>conjecture
>reddit posts

give me something directly from AMD if you want to claim that "beating skylake" was one of their allegations

how can you make a proper approach when they're ES samples?

the fact we had one ES sample that was LOWER than bulldozer and then another that was barely faster than a haswell i5, who is there to say there isn't another one thats faster than it that wasn't shown?

WE CANNOT draw any conclusions because ultimately we don't know fucking shit about any of AMD's ES samples.

we don't have a clue how they're setup or how the benchmarks were even tested.

we don't know shit.

all these articles keep doing is spreading fud and spewing nonsense from people.

In both Canada and UK it's still roughly the same price on Newegg, taxes and VAT not included.

It'll be sad if Zen fails though, as that'll probably be when AMD decides to ax their entire CPU division.

ES chips are often clocked lower than their retail equivalents because they're used to test functionality not performance. They might not even be fully enabled for all we know they have disabled specific instruction set extensions or disabled turbo. The software can be configured to ignore certain features as well

Who is to say god won't destroy the world tomorrow? HUH? HUH? PRAY TO GOD TODAY.

In a rational world of business, people will naturally go for what's the most efficient. Given two different engineering samples, the logical conclusion would lead to AMD adopting the higher performing design over lower performing design. This is not rocket science.

If you had to choose between 10 dollar and 5 dollar, everyone would choose $10 on high probability chance. Barring the retards.

So when there are two design frameworks, AMD would naturally get the best performing one and reprint that shit. There's no "well AMD might choose the lower performing ES design over higher performing ES design" scenario in real world scenarios.

>as that'll probably be when AMD decides to ax their entire CPU division.

not bloody likely. they have some highly lucrative contracts with console vendors and also selling SoCs in China.

worst case scenario I expect it'll be their last x86 desktop product line but they'll definitely keep making CPUs for a long time.