Coffeelake thread, benchmarks, reviews, input, everything.

[raghu78]

Yes i can, because it's true. The difference is in the process like i said. Coffeelake is just 2 cores added on to the 4 existing Kabylake cores, i don't consider that a spectacular engineering achievement at all. It's and engineering effort yes, but certainly not spectacular achievement. If it was a 50% IPC improvement, that would be spectacular. Adding 50% more cores doesn't translate into 50% more processing capability.
The TDP doesn't align with the power draw of the chip, that is special condition number that intel seems to have tagged on for marketing purposes. Is it 50% more power for 50% more cores? No it isn't, because the process improved.
This is a pivotal moment in the industry. In some respects, intel has now lost that process lead. A physics wall is fast approaching and any remaining lead is going to diminish to nearly nothing. This is where the rubber is going to meet the road and who has the best architecture for scalability. I think intel has unfortunately relied too long on it's process for it's dominance in the industry and forward thinking on design has taken a back seat.

To the poster that claimed Coffeelake wasn't a paper launch, it is the exact definition of one, not sure how you can say otherwise. This is a trickle of product to the market in ultra low volume.

I can say this no company on earth including Intel, Apple or AMD can improve IPC by 50% on their existing high performance CPU core. Getting 10% improvement is considered very good. Apple manages to get 20% IPC growth with a very tight integration of CPU chip design,OS and first party apps which make good use of any of the architectural improvements and/or ISA extensions.

Yes, so what it boils down to is a 10% IPC lead, which subsequent versions of Zen should eliminate, and a process that clocks higher. So at clock parity, the 8700K would have a 10% lead in stricly single threaded applications (of which there are very few) and fall well short in performance for multi threaded applications compared to the uncut 8 core Ryzen chip at the same price point. As i said in the other post, yes it is an engineering effort but not spectacular. The process allows it to clock higher.

Neither Intel nor AMD is standing still. Both companies continue to make architectural improvements. Icelake will bring IPC improvements to Skylake as Zen 2 will bring to Zen. How much AMD is able to narrow the IPC gap or will the gap widen will be known when the products launch in 2019. As for process related clock speed advantage that is a key advantage for Intel . GF and AMD need to be able to compete on that front too if they want to compete on overall performance (ST and MT). The fact that the 8700k is competitive with 1800x for MT while crushing it on ST points to the significant gaps AMD has to cover both in IPC and clock speed.

 

[PeterScott]

I can say this no company on earth including Intel, Apple or AMD can improve IPC by 50% on their existing high performance CPU core. Getting 10% improvement is considered very good. Apple manages to get 20% IPC growth with a very tight integration of CPU chip design,OS and first party apps which make good use of any of the architectural improvements and/or ISA extensions.

Also Apple had big jumps in IPC because they were following already broken in trail. Essentially they are adding features that Desktop CPUs have had for years, once Apple attains feature parity, it's gains will grind down as well.

It is much harder being on the leading edge of desktop performance, breaking the new ground.

 

[raghu78]

Also Apple had big jumps in IPC because they were following already broken in trail. Essentially they are adding features that Desktop CPUs have had for years, once Apple attains feature parity, it's gains will grind down as well.

It is much harder being on the leading edge of desktop performance, breaking the new ground.

I cannot say thats entirely true. Apple is bringing unreal performance to the smartphones. The 20% ST perf jump from A10 to A11 Bionic comes at almost same clock speed. Whats amazing is the A10 was already a monster in IPC and ST performance thrashing the Android competition by >50% and competing well with Intel Core M chips. Right now A11 Bionic is competing with 15w Intel chips. The advantage for Apple is they don't need their CPU design to clock > 4 Ghz and its enough if they hit 2.4 Ghz whereas Intel's core has to scale through a wide clock range from 2Ghz to 4.5 Ghz. Apple's CPU uses higher density libraries and shorter cells than Intel does given their different design goals.

 

[PeterScott]

I cannot say thats entirely true. Apple is bringing unreal performance to the smartphones. The 20% ST perf jump from A10 to A11 Bionic comes at almost same clock speed. Whats amazing is the A10 was already a monster in IPC and ST performance thrashing the Android competition by >50% and competing well with Intel Core M chips. Right now A11 Bionic is competing with 15w Intel chips. The advantage for Apple is they don't need their CPU design to clock > 4 Ghz and its enough if they hit 2.4 Ghz whereas Intel's core has to scale through a wide clock range from 2Ghz to 4.5 Ghz. Apple's CPU uses higher density libraries and shorter cells than Intel does given their different design goals.

None of that counters my argument. Apple is more aggressive bringing high end desktop CPU features to mobile than it's competition. Greater depth in multi-issue capabilities. But this quickly hits a wall, because each increase drives much greater complexity in branch prediction, and bigger in flight caches.

Apple may have now caught up to Intel in this capability and the next leap forward may be extremely costly (which is why Intel has appeared stagnant for years).

The trajectories of the last 5 years are not going to continue for the next 5. 5 Years from now I expect Apple/Intel/ARM competitors to be very close, as they are all much closer to the wall on diminishing returns for the architecture advancement.

 

[DrMrLordX]

Apple may have now caught up to Intel in this capability and the next leap forward may be extremely costly (which is why Intel has appeared stagnant for years).

We won't know until A12/A12x shows up. I don't think many people expected what A11x's "fast" cores brought to the table. Regardless, there's a separate thread or two for that . . .

 

[Excessi0n]

Yeah, they should show minimum framerates or something because the gaming benchmarks look the same as the sdram cl2 versus cl3 benchmarks from 15-20 years ago. "Fast RAM makes a big difference!*"

*Assuming 3% counts as big.

Minimums? Nah. They should be showing actual frametime charts. And testing games which are actually CPU-intensive. Where are the benchmarks of a jog through downtown Boston in Fallout 4, or the launch of an idiotically huge rocket in KSP?

 

[DrMrLordX]

Fallout 4 isn't "CPU-intensive". Fallout 4 is just flat-out badly-designed. Ditto for Starcraft II.

 

[TheF34RChannel]

Fallout 4 isn't "CPU-intensive". Fallout 4 is just flat-out badly-designed. Ditto for Starcraft II.

It's a graphic disgrace!

Anyway, they need to test CPU bound games alright instead of mere popular titles. For example, the Witcher 3, watch dogs 2 and what not.

 

[FIVR]

The trajectories of the last 5 years are not going to continue for the next 5. 5 Years from now I expect Apple/Intel/ARM competitors to be very close, as they are all much closer to the wall on diminishing returns for the architecture advancement.

I think this is not an accurate prediction. Even if you only consider apple engineers to be "the equal" of intel engineers (which i think most here would see as overly generous to intel), Apple is projected to surpass 1T market cap and will be able to buy and sell an intel-sized company with ease. They can easily spend 10x what intel can on engineering and process technology, if they want. Even if only look at the pure economics of the situation... unless you believe that investment in technology is a fruitless endeavor (which I think most here would know is wrong) - it is obvious that Apple will dominate all competitions in the semiconductor space in the very near future.

 

[piesquared]

I can say this no company on earth including Intel, Apple or AMD can improve IPC by 50% on their existing high performance CPU core. Getting 10% improvement is considered very good. Apple manages to get 20% IPC growth with a very tight integration of CPU chip design,OS and first party apps which make good use of any of the architectural improvements and/or ISA extensions.



Neither Intel nor AMD is standing still. Both companies continue to make architectural improvements. Icelake will bring IPC improvements to Skylake as Zen 2 will bring to Zen. How much AMD is able to narrow the IPC gap or will the gap widen will be known when the products launch in 2019. As for process related clock speed advantage that is a key advantage for Intel . GF and AMD need to be able to compete on that front too if they want to compete on overall performance (ST and MT). The fact that the 8700k is competitive with 1800x for MT while crushing it on ST points to the significant gaps AMD has to cover both in IPC and clock speed.

Don't agree.It has been done, whether regardless of where the starting point is and it is remarkable to do it in 1 generation. I believe intel claimed a 15% increase? You are saying they achieved 10. I think reviews show closer to 0 though so who to believe. 10% hasn't ever been considered very good as far as i can remember. I know many people claimed intel's incremental increases for the past several years to be meager.

If intel ever gets it's 10nm process ready to roll, i think i remember reading that it is expected to be a lower performance process than 14++. It has been reported what performance increase to expect from GF's 12nm. So at the very least we can expect either 12nm vs 14++, and those clock speeds on 14++ are very unlikely to increase, and we can expect a 12nm vs 10nm late next year where the 12nm process is likely to exceed intel's 10nm process. I think it's safe to assume Zen2 will come with an IPC uplift, not sure what to expect from intel's next gen but that will be at least a year after Coffeelake. So in the short term we will have existing Coffeelake on 14++ vs Zen on 12nm with slight improvements to the core is my guess. Either way AMD will gain on IPC and clockspeed vs existing Coffeelake cores.

 

[DrMrLordX]

I think this is not an accurate prediction. Even if you only consider apple engineers to be "the equal" of intel engineers (which i think most here would see as overly generous to intel), Apple is projected to surpass 1T market cap and will be able to buy and sell an intel-sized company with ease. They can easily spend 10x what intel can on engineering and process technology, if they want. Even if only look at the pure economics of the situation... unless you believe that investment in technology is a fruitless endeavor (which I think most here would know is wrong) - it is obvious that Apple will dominate all competitions in the semiconductor space in the very near future.

That depends on how Apple chooses to spend their money. If they are as aggressive in R&D as a company like Intel, then yes, they can easily outspend Intel and bury them on the design front. There are already signs that Apple is doing just that, albeit "from behind". They haven't caught up yet, and they're still in a niche (albeit a large and highly-competitive niche). Apple isn't selling full laptop, desktop, or server processors from their own design house. Yet.

Apple is still mostly a design-and-software company. They are slowly replacing hardware from outside vendors with internally-designed stuff in their mobile products. They may well do that eventually with their laptop, desktop, and workstation/server products. And they have an integrated software ecosystem to go right along with it. BUT! But. If they go that route, the Antitrust Furies (no, not furries, lulz) may well be on their heels.

 

[Bouowmx]

I believe intel claimed a 15% increase?

https://s21.q4cdn.com/600692695/fil...2017_Intel_Investor_Meeting_Renduchintala.pdf (PDF page 14)
15% increase in performance of one SYSmark sub-test. How does one interpret this claim to be IPC?

If intel ever gets it's 10nm process ready to roll, i think i remember reading that it is expected to be a lower performance process than 14++.

https://newsroom.intel.com/newsroom.../2017/03/Kaizad-Mistry-2017-Manufacturing.pdf (PDF page 29)
What is "transistor performance"? Is it just frequency? Misc.: Remember 10 nm is mobile-only (Cannon Lake).

 

[coercitiv]

https://s21.q4cdn.com/600692695/fil...2017_Intel_Investor_Meeting_Renduchintala.pdf (PDF page 14)
15% increase in performance of one SYSmark sub-test. How does one interpret this claim to be IPC?

Friendly reminder to anyone who saw the 7th -> 8th gen 15% perf increase claim that it was based on comparing 15W TDP 2c/4t versus the new 15W TDP 4c/8t Kaby Lake CPUs.

The only thing that might bump IPC a bit in the new Kaby Lake R is the 8MB L3 cache. Previously the ULV parts had 4MB cache, and even many of the 45W TDP 4c/8t had only 6 MB L3.

 

[Dayman1225]

http://browser.geekbench.com/v4/cpu/4431108
http://browser.geekbench.com/v4/cpu/4443759

What looks to be Coffee lake 6 core mobile spotted on geekbench

Base Frequency 2.39 GHz
Maximum Frequency 3.59 GHz

Since when did GB report max freq?

 

[Lodix]

http://browser.geekbench.com/v4/cpu/4431108
http://browser.geekbench.com/v4/cpu/4443759

What looks to be Coffee lake 6 core mobile spotted on geekbench

Base Frequency 2.39 GHz
Maximum Frequency 3.59 GHz

Since when did GB report max freq?

It got updated recently.

 

[jpiniero]

Frequency seems kinda low for what in theory would be an i7 45W H. Figure the turbo would be at least 4 Ghz.

 

[Lodix]

Frequency seems kinda low for what in theory would be an i7 45W H. Figure the turbo would be at least 4 Ghz.

Max all core turbo ?

 

[jpiniero]

Max all core turbo ?

No, I was talking about single. KBL-R has a SCT of 4 or 4.2; so you figure that 45W would be able to do at least that; no?

 

[Lodix]

No, I was talking about single. KBL-R has a SCT of 4 or 4.2; so you figure that 45W would be able to do at least that; no?

I am not aware of all the products from Intel but I was implying that maybe the number was for all core turbo. But yeah single core on 14nm++ should be higher than past kabylake products of the same TDP.

 

[Dayman1225]

Intel i5-8400 Review: 2666MHz & 3200MHz Gaming Benchmarks Gamers Nexus

Article found here

 

[PeterScott]

Intel i5-8400 Review: 2666MHz & 3200MHz Gaming Benchmarks Gamers Nexus

Article found here

Nice little RAM boost.

I don't really agree that it makes no sense to get a Z370 MB for a Locked processor. It isn't like they are $300 MB. You also get more features with the Z MBs.

I want S/PDIF sound, and it tends to be missing on cheaper MBs.

 

[hnizdo]

Fallout 4 isn't "CPU-intensive". Fallout 4 is just flat-out badly-designed. Ditto for Starcraft II.

I love these proclamation coming from "forum experts". Fallout 4 is complex game, of course its cpu intensive.

 

[DrMrLordX]

I love these proclamation coming from "forum experts". Fallout 4 is complex game, of course its cpu intensive.

Who said anything about being a forum expert? Fallout 4's engine is an ancient retread of the same damn engine they used years ago in Fallout 3. It's a hacked-together mess. Using it as a shining example of anything, whether CPU-intensiveness or otherwise, is a mistake.

Next you'll be lecturing us all about how Starcraft II is a "CPU-intensive" game.

 

[hnizdo]

Fallout 4's engine is an ancient retread of the same damn engine they used years ago in Fallout 3. It's a hacked-together mess.

You are missing a point. Practically all current game engines are more or less descendants of some ancient codes. It doesnt matter at all if its working. The only thing that matters is that's largely played AAA game and its CPU demanding. Its perfectly OK to make it as reference for CPU benchmarking.

Nice example is AotS. AotS is trash game played by a few masochist developed for AMD as a tech demo. Despite that fact its still included in game benchmarks as "shining" example of DX12 engines.

 

[krumme]

Intel i5-8400 Review: 2666MHz & 3200MHz Gaming Benchmarks Gamers Nexus

Article found here

What is the cas settings for 2666? Its c16 for 3200.
My point beeing if its same cas its kind of misleading test. You can eg run the 8400 on a non oc z board on 2666 c14 easily. You can tweak latency and settings not bandwith.
Steve knows that and to be fair it irritates me he just dint flat out write the latency for the 2666 ram but gives a non link to prior cpu reviews. It comes off a bit fishy.

In more general view many games is latency dependant and thats why we see ryzen perform worse. Especially with slow c16 settings. I simply dont agree in comparing platforms like that. Ryzen needs low latency and a 8700k a good cooler so it dont throttle. If you dont give them that its imo biased testing.

In this the purpose seems to be look 3200 is much better for a 8400 go get you 370 but imo it comes of as more as a sales piece. I need some c14 2666 to confirm.
I am pretty sure the results is more or less excactly the same.