AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[mikk]

Except what new CPU could they release? (That's not Skylake-X or Kabylake-X) That's sort of why I think it's BS.

This is for sure BS.

 

[KTE]

What do you believe Intel will respond with?

I'm not privy so I can only presume based on historical data and common business sense.

IF AMD is competitive top down, in absolute performance;

Intel will never sit back while AMD attacks their most profitable SKUs. They have billions to defend their share.

Intel will already have modeled Ryzen and their response ever since the chip architecture and hype was known.

Intel will also already have been preparing their immediate counter. If Ryzen launches in March, that would be for 12 months minimum.

Intel will accelerate their next chip roll-out. As was done with Penryn when Barcelona launched.

Intel will try to wreck AMDs launch and take away the limelight.

Intel will release higher frequency models.

Intel will slash prices.

Intel will release more flexible and lower power SKUs.

AMD literally has K2 to climb. Making sure they are near parity in performance is only the start for them.

It is entirely plausible that AMD would release a chip 5% below BDe/SKL in IPC, and due to all the above, still not gain much reputation or share.

Sent from HTC 10
(Opinions are own)

 

[Glo.]

Intel will release higher frequency models.

Intel will slash prices.

Intel will release more flexible and lower power SKUs.

AMD literally has K2 to climb. Making sure they are near parity in performance is only the start for them.
(Opinions are own)

I will put some things into perspective. 3.4/3.7 GHz, 8 core/16 Thread CPU from AMD costs 399$, and offers slightly higher performance than 6900K, that costs 1099$. What is Intel's answer, in your opinion?

Skylake-X? Kaby Lake Price cuts? Nothing looks competitive in that situation, with AMD. Who has in the end K2 to climb?

 

[majord]

It's very very likely not going to offer slightly higher perf than a 6900K though. rather slightly less performance.

I know people are getting edgy so close to release, but it seems like Ryzens IPC and throughput keeps mysteriously increasing every day for no apparent reason. With another month till launch I fear that when it crashes back to reality on release day, What is actually an excellent product, and for that matter a stellar achievement for AMD, will be marred by both disappointed dreamers venting frustration, and trolls taking advantage of the situation.

I think people have lost a bit of perspective here.. These rumors of Intel reacting probably are true, but Ryzen doesn't have to be faster for that to be so. Simply being within 10, even 15% of Intel's Halo products at such [rumoured] competitive prices is enough.

There is a chance performance will be higher than this of course, But why not let that remain a 'chance' and not an expectation.

 

[Sven_eng]

I think most people expect lower than Broadwell-E IPC, but XFR looks like a wildcard.

Few people seem to be talking about clock speeds on 4C and 6C parts. If AMD can do 8C/16T at 3.6/4GHz in under 95W, they can go a lot higher with 4C/8T. We haven't seen any benchmarks with boost yet nevermind XFR.

 

[Glo.]

It's very very likely not going to offer slightly higher perf than a 6900K though. rather slightly less performance.

Yeah, and in every benchmark demoed by AMD in New Horizon event, Ryzen at 3.45 GHz was what? Slower than 6900K? Or slightly faster.

And suddenly it has to be slower with 3.7 GHz boost clock?

 

[Arachnotronic]

I thought kabylake was already super awesome and not as disappointing as everyone wanted us to believe?

Kaby Lake is awesome. I have a couple of 7700Ks here and even the "bad" one clocks better than the best 6700K I ever had (4.9GHz).

The good one does 5GHz with 4.7GHz uncore, no AVX negative offset.

Kaby Lake is good stuff, but if Intel is pushed to put out faster CPUs (higher stock clocks, higher core count CPUs, etc.), then I am a happy camper. When Intel is pushed to bring out their best, they can deliver some seriously awesome stuff.

 

[Arachnotronic]

Yeah, and in every benchmark demoed by AMD in New Horizon event, Ryzen at 3.45 GHz was what? Slower than 6900K? Or slightly faster.

And suddenly it has to be slower with 3.7 GHz boost clock?

At the New Horizon event AMD showed a grand total of two benchmarks. Considering that this is a marketing event, you would expect AMD to pick the cases that put its new baby in the best light.

 

[inf64]

At the New Horizon event AMD showed a grand total of two benchmarks. Considering that this is a marketing event, you would expect AMD to pick the cases that put its new baby in the best light.

That is indeed true but those two tests are different workloads that stress different parts of the core, both showing similar trend. I think that Ryzen is AMD's Haswell done on 14nm for all intents and purposes. I expect similar IPC and due to process and design choices better perf./watt. Since BDW-E does have slightly higher IPC across the board (according to anandtech around 3.3%) and Skylake brings another 3% or so, Ryzen will have lower IPC than SKL- I expect 7-10% deficit.

Also because KL really does clock so high (5GHz), I think intel will definitely have some 20-25% higher aggregate ST performance when both Ryzen and KL are pushed to the max- 4.3Ghz for 4C/6C/8C Ryzen Vs 5Ghz KL.

So there is a niche within a niche market and those who would want absolute best ST performance will undoubtedly go KL route.

 

[CentroX]

I am even more confused today. It is hard to know whats reality and whats fiction.

 

[JDG1980]

If Zen threatens Intel's dominance, their most obvious countermove would be to speed up the release of Coffee Lake. Intel's edge in maximum clock speed and IPC should be enough for 6C/12T Coffee Lake to be reasonably competitive with 8C/16T Zen. Current rumors have Coffee Lake not coming until 2018, but it's apparently just a hex-core variant of Kaby Lake on the same 14nm process, so there's no particular reason that Intel couldn't move up the timeline if they had to.

In the meantime, Intel could institute modest price cuts on its Kaby Lake lineup, and refresh it with higher clock speeds (there's still some headroom left, and replacing TIM with solder should be good for a couple hundred MHz). But if Zen is what the optimistic leaks indicate it might be, then 4C/8T Kaby Lake will have trouble competing with 8C/16T Zen in all but the most lightly threaded applications.

 

[Arachnotronic]

If Zen threatens Intel's dominance, their most obvious countermove would be to speed up the release of Coffee Lake. Intel's edge in maximum clock speed and IPC should be enough for 6C/12T Coffee Lake to be reasonably competitive with 8C/16T Zen. Current rumors have Coffee Lake not coming until 2018, but it's apparently just a hex-core variant of Kaby Lake on the same 14nm process, so there's no particular reason that Intel couldn't move up the timeline if they had to.

In the meantime, Intel could institute modest price cuts on its Kaby Lake lineup, and refresh it with higher clock speeds (there's still some headroom left, and replacing TIM with solder should be good for a couple hundred MHz). But if Zen is what the optimistic leaks indicate it might be, then 4C/8T Kaby Lake will have trouble competing with 8C/16T Zen in all but the most lightly threaded applications.

It seems that the entire internet has completely forgotten about the existence of Intel's high end desktop platform, which has six, eight, and 10 core options.

As to your post, what makes you think that Intel can just "speed up" a chip release? It's not as though Intel completes a chip design and then sits on it for so long that they can just release it much earlier if need be.

 

[coercitiv]

As to your post, what makes you think that Intel can just "speed up" a chip release?

Based on early info from CanardPC they are definitely "speeding up" something:

Tiens, Intel commence à flipper au sujet de Ryzen et sample en catastrophe de nouveaux CPU. On vous en dit plus lundi (sample en route).

However, it all seems odd to me, since the (twitter) discussion orbits around the idea of Intel increasing core count in Broadwell-E, which in my humble view is a rather poor approach: Intel already has core count advantage, what they need is increased clocks, hence lower power usage... hence 14nm+. BDW-E on 14nm+ makes no sense with SKL-X so close, so this either leaves some type of improvements for KBL to push frequency even higher (ugly diminishing returns) or alternatively very aggressive schedule for SKL-X.

My bet is the latter, every month spent with Zen fighting BDW-E instead of SKL-X is a month of free advertising for AMD products. I guess we'll find out tomorrow.

 

[DrMrLordX]

I thought kabylake was already super awesome and not as disappointing as everyone wanted us to believe?

Nah, it's just that Intel hasn't launched anything that good with more than 4 cores (yet). Most folks are cynically expecting Broadwell-E-like clockspeeds when Skylake-X comes out. I'm not so sure that will be the case, but whatevs.

I am even more confused today. It is hard to know whats reality and whats fiction.

Keep going with that feeling. It's the closest thing to the truth . . . in general.

It seems that the entire internet has completely forgotten about the existence of Intel's high end desktop platform, which has six, eight, and 10 core options.

Well no, it's just that those things are pretty expensive, overall. Maybe not the 6c chips so much, but everything else? woof.

 

[Doom2pro]

What benchmark would have caused Intel to panic though? I haven't seen one.

Mind you I'd like to see Intel release a quick response to RyZen. This sort of competition is what we've all been waiting for.

I don't think it's any benchmark we have seen, I think they got their hands on it somehow, either directly or indirectly to see how good it really is so they can get a jump start on a response and it looks like they got a bell ringer.

 

[inf64]

If I had to guess what it is, I will put my bet on a 8C/16T BDW-E part with higher base and Turbo clock than 6900K . Then they could say look, this is the new top 16T model we have, now we can cut the price of the old one to ~700-800$. I could see intel releasing 6930K model with 3.4-3.5Ghz base and 4Ghz boost speed and pricing it at 1K$.

This way they could probably beat top Ryzen by a small margin and claim they still have top 8C/16T performance and ask a premium for it. This would kinda suck for SKL-X though since it will have a hard time beating such highly clocked part.

 

[Arachnotronic]

Well no, it's just that those things are pretty expensive, overall. Maybe not the 6c chips so much, but everything else? woof.

We have pretty much zero information or even leaks about AMD's Ryzen pricing.

 

[Doom2pro]

We have pretty much zero information or even leaks about AMD's Ryzen pricing.

We have something even better! A mountain of speculation!

 

[Arachnotronic]

Based on early info from CanardPC they are definitely "speeding up" something:

However, it all seems odd to me, since the (twitter) discussion orbits around the idea of Intel increasing core count in Broadwell-E, which in my humble view is a rather poor approach: Intel already has core count advantage, what they need is increased clocks, hence lower power usage... hence 14nm+. BDW-E on 14nm+ makes no sense with SKL-X so close, so this either leaves some type of improvements for KBL to push frequency even higher (ugly diminishing returns) or alternatively very aggressive schedule for SKL-X.

My bet is the latter, every month spent with Zen fighting BDW-E instead of SKL-X is a month of free advertising for AMD products. I guess we'll find out tomorrow.

Nah, Broadwell-E on 14nm+ isn't happening, that would require significant design work and the product wouldn't be in the market for very long.

My personal guess -- for whatever it's worth -- is that this will be a higher clocked Kaby Lake-S SKU. It's still too early for X299 (all of the mobo makers just launched their Z270 designs and they aren't going to want the headache of putting out a bunch of X299 designs so soon afterwards) and Broadwell-E silicon is long finished. They could, say, unlock the multiplier on the Xeon E5 2687v4 and sell it at the 6950X's price point, then push the other Broadwell-E parts downwards in pricing, but that seems pretty messy.

 

[lopri]

AMD has been shouting about Ryzen for a while now, and releasing a higher clocked Kaby isn't exactly hard. It means more aggressive binning (testing for and finding samples that can hit the desired frequencies at the desired power consumption), but no fundamental silicon changes.

Given that all Kabys seem to be able to hit 4.8ghz+, I think Intel has a good deal of headroom to put out faster SKUs out of the box. Won't matter so much to people who will OC the chips anyway, but reviews tend to test stock performance.

That is what I think as well. Actually I have thought like half of Intel's manufacturing prowess is in the binning, especially since they hit the wall after SB. Base freq., Turbo freq., Max Turbo freq., All-Core Turbo freq., AVX-enabled Turbo freq., etc. etc. Binning and savaging has to be their No.1 expertise. It should not be hard for them to release a new SKU or two with cherry-picked dies.

 

[JDG1980]

It seems that the entire internet has completely forgotten about the existence of Intel's high end desktop platform, which has six, eight, and 10 core options.

Intel's HEDT platform will likely still be able to beat AMD on raw performance, but it will be difficult for it to beat Ryzen in performance per dollar in the midrange. The motherboard design makes it inherently more expensive than LGA 11xx and AM4.

As to your post, what makes you think that Intel can just "speed up" a chip release? It's not as though Intel completes a chip design and then sits on it for so long that they can just release it much earlier if need be.

I think that's exactly what they have been doing, due to the lack of any meaningful competition in x86 in the past couple of years. Intel could release Coffee Lake in 6 months if they wanted to.

 

[lopri]

^ Yup. Kabylake, my a**. More like a new stepping Skylake. Or rather, Skylake with a new firmware.

 

[DrMrLordX]

We have something even better! A mountain of speculation!

There have been a few people swaggering around claiming to have insider info that the top-end Ryzen will be $499, will max OC to 4.1-4.2 GHz, and will launch Feb 28th. And really AMD would be insane to price their top-end chip higher than that anyway . . .

 

[zinfamous]

It seems that the entire internet has completely forgotten about the existence of Intel's high end desktop platform, which has six, eight, and 10 core options.

As to your post, what makes you think that Intel can just "speed up" a chip release? It's not as though Intel completes a chip design and then sits on it for so long that they can just release it much earlier if need be.

Easy to forget about $800-1k chips when it comes to the mainstream market ("the entire internet"), but then that's kinda the point of AMD targeting the real mainstream market with ~mainstream-priced HEDT competitive chips.

 

[Arachnotronic]

Easy to forget about $800-1k chips when it comes to the mainstream market ("the entire internet"), but then that's kinda the point of AMD targeting the real mainstream market with ~mainstream-priced HEDT competitive chips.

AMD has referred to Summit Ridge as targeting the high end desktop market.

I have no knowledge of AMD's pricing plans, but the implicit assumption that AMD is going to offer 6900K features/performance for 7700K pricing appears unfounded in light of the company's previous behavior (see: FX 9590 pricing at launch).