Discussion Intel current and future Lakes & Rapids thread

[shady28]

The power consumption would have been bad, but an 11900k would have been better than the 9900ks for the intended target market.

Rocket Lake has already been banished from the mobile space, and Intel isn't launching any 4c-or-smaller variants either. It's solely an enthusiast's part for the DiY sector. With an 8c limit, you know it isn't for "serious" productivity workloads either.

No to feed your trolling habits, but Rocket Lake with its 11XXXT version is clearly not "solely an enthusiast's part for the DiY sector." And the first leaks were on an HP Omen, i.e. a major OEM system, something Zen 3 has yet to be seen on.

 

[DrMrLordX]

No to feed your trolling habits

Sure pal, sure.

but Rocket Lake with its 11XXXT version is clearly not "solely an enthusiast's part for the DiY sector." And the first leaks were on an HP Omen, i.e. a major OEM system, something Zen 3 has yet to be seen on.

HP has a relationship with Intel that it will maintain for some time yet. The same applies to other OEMs. Don't conclude that "well it's in the HP Omen hurr durr" is somehow a compelling argument for using the 11900K over a 5950X in productivity workloads. The only way people will get Rocket Lake-S in their corporate desktops is if procurement has a contract with a vendor that'll make it happen. An 11900k might also be faster in some CAD work. Might.

Otherwise, nobody, and I mean nobody, with an ounce of common sense would use a 11900K in a 1P workstation for doing actual work. AMD has all the important MT benchmark wins in rendering, software encoding, compiling, you name it, versus Intel's lineup. The 5950X even threatens the 10980XE in productivity workloads.

 

[tamz_msc]

Sure pal, sure.



HP has a relationship with Intel that it will maintain for some time yet. The same applies to other OEMs. Don't conclude that "well it's in the HP Omen hurr durr" is somehow a compelling argument for using the 11900K over a 5950X in productivity workloads. The only way people will get Rocket Lake-S in their corporate desktops is if procurement has a contract with a vendor that'll make it happen. An 11900k might also be faster in some CAD work. Might.

Otherwise, nobody, and I mean nobody, with an ounce of common sense would use a 11900K in a 1P workstation for doing actual work. AMD has all the important MT benchmark wins in rendering, software encoding, compiling, you name it, versus Intel's lineup. The 5950X even threatens the 10980XE in productivity workloads.

Intel MKL says hi.

 

[dmens]

The power consumption would have been bad, but an 11900k would have been better than the 9900ks for the intended target market.

Rocket Lake has already been banished from the mobile space, and Intel isn't launching any 4c-or-smaller variants either. It's solely an enthusiast's part for the DiY sector. With an 8c limit, you know it isn't for "serious" productivity workloads either.

That intended target market is so vanishingly small that I do not even consider it a worthwhile market for engineers to target directly. The only reason it is being pumped up by the Intel marketers is because they don't have any talking points left.

 

[Hulk]

Krzanich's major strategic mistake was not doing that backport to 14nm starting in 2017/2018 when they knew 10nm might be delayed. In fact, they probably should have done it whether they were going to be delayed or not. It's called a fall-back plan, hedging your bets, an "insurance policy".

Apple has been porting Darwin / MacOS / OS X to ARM for years in its labs, "just in case". Apple did the same thing back in the days of PowerPC, keeping an x86 port handy. A few hundred million to do that each year for a company like Intel would have been nothing.

Now they are racing against the clock to see if they can get Rocket Lake out before AMD can get real supply of Zen 3 into the market, all AMD really has right now is bragging rights from some chip sales via NewEgg or mindshare.de to about 2% of the market. I figure Intel has about 3 months. Rocket Lake has to be good.

Exactly.

While the process was getting worked out they could have continued with architectural advancements. Today they could have been porting a core 3 generations ahead of Willow Cove to 14nm. In fairness they did make some sizeable gains with the iGPU. While I realize it's still weak for gaming for guys like me who need OpenCL for video editing it will be a nice boost.

 

[DrMrLordX]

Intel MKL says hi.

That situation seems to change month-by-month. Someone creates a workaround for MATLab on AMD CPUs, and then it breaks again, and then it's un-broken, etc.

That intended target market is so vanishingly small that I do not even consider it a worthwhile market for engineers to target directly. The only reason it is being pumped up by the Intel marketers is because they don't have any talking points left.

Gaming is a thing. Plus marketing is what Intel does best right now. If they can keep some mindshare by winning some game benchmarks, well, what else do you expect them to do?

 

[tamz_msc]

That situation seems to change month-by-month. Someone creates a workaround for MATLab on AMD CPUs, and then it breaks again, and then it's un-broken, etc.

MATLAB is but one software package that uses MKL. There's Anaconda, Mathematica, ANSYS etc. The situation with MKL performance on Zen is fluid. Some things work quite well while others do not and it changes all the time. As long as this is the case Intel is a viable option for workstation use. Not all use-cases require 16 cores of the 5950X.

 

[mikk]

MSI Prestige 14 Evo

The MSI Prestige 14 Evo Review: Testing The Waters Of Tiger Lake

www.anandtech.com

Razer Blade Stealth 13 vs Lenovo Slim 7

The Razer Blade Stealth is one of the few current Tigerlake designs with a good cooling setup and LPDDR4x-4266. Similar to the MSI Prestige 14 Evo iGPU gaming is easily faster in most games than the best Renoir-U.

 

[dmens]

Gaming is a thing. Plus marketing is what Intel does best right now. If they can keep some mindshare by winning some game benchmarks, well, what else do you expect them to do?

Couldn't care less what marketing does. What the engineering teams could have done is come up with something better than Skylake+++. For those who believe Intel halted CPU development due to 10nm delays, that is wrong, they kept going but all they could come up with was more Skylake.

 

[Bam360]

Couldn't care less what marketing does. What the engineering teams could have done is come up with something better than Skylake+++. For those who believe Intel halted CPU development due to 10nm delays, that is wrong, they kept going but all they could come up with was more Skylake.

They didn't halt development of architectures, they just designed these futures architectures with 10nm in mind. What they should've done is design a better architecture for 14nm instead of refreshing Skylake, that's it. If they couldn't do this, then they are incompetent as a designers of architectures, seeing as other companies are capable of doing good designs on the same node.

 

[inf64]

Videocards has an article about Rocketlake ES that was tested and results were posted on bilibili : https://videocardz.com/newz/intel-c...engineering-sample-tested-on-b560-motherboard

Summary of R15 and R20 results VS Skylake and Zen3:

Rocketlake 8C/16T (base 1.8Ghz, all core turbo 3.8Ghz, ST turbo 4.4GHz)
R15 ST 217pts
R15 MT 1929pts

R20 ST 529pts
R20 MT 4683pts


Skylake 8C/16T ( base 3.6GHz, all core turbo 4.7GHz, ST Turbo 5Ghz); bilibili results were lowish for 9900K so I used guru3ds as it matches better with 10900K and 10700
R15 ST 216
R15 MT 2044

R20 ST 512
R20 MT 4853

5800X 8C/16T ( base 3.6GHz, ST Turbo 4.7Ghz), source for the scores
R15 ST 266
R15 MT 2639

R20 ST 624
R20 MT 6112

ST IPC comparison (same clock) in R15 and R20 relative to Skylake (100% or 1):

R15
Skylake 1
Cypress Cove 1.14
Zen3 1.28 (5800X might be boosting to 4.8Ghz as many reviews show so I used 4.8Ghz)

R20
Skylake 1
Cypress Cove 1.174
Zen3 1.25 (5800X might be boosting to 4.8Ghz as many reviews show so I used 4.8Ghz)

In conclusion, R15 shows 12% IPC lead for Zen3 while newer R20 shows 6.4%. Basically there will be benchmarks where Cypress Cove might perform like Icelake/Willow Cove and there will be cases where it won't. If it clocks to 4.7Ghz all core turbo it might match 5800X or very slightly edge it out in some games. It will need 5.3Ghz ST Turbo to come close to 5800X ST performance

 

[shady28]

This is a new leak of a 11900 non-K ES, showing CPU-Z benchmark scores.

Intel Core i9-11900 8 Core Rocket Lake Desktop CPU Benchmarks Leak Out, Early ES Chip With 1.8 GHz Clocks Tested on Z490 Platform

Intel's Core i9-11900 8 Core Rocket Lake Desktop CPU benchmarks have leaked out, showing some impressive IPC gains for the 11th Gen family.

wccftech.com

The CPU has a 1.8/4.4 GHz Base/Boost clock-speed and the IPC boost in this particular benchmark is roughly ~20%.

Looking at those benchmarks from different sites, it looks like everything from the 11700/11700T/11700K and up is only distinguished by its TDP rating (also related to binning), base/single/multi-core multipliers, and locked/unlocked.

The source article, translated, is also revealing. From that translation :

1 - This might not even be a 10900, they are assuming that.

2 - This is a very low end motherboard, and may not even be a 5-series :



3 - The clock speeds actually correspond to an 11700T, a SKU that based on 11700T I would expect to find in high end AIO desktop builds and with mini-PCs from OEMs :




4 - This chip is clocked below the expected single and multi-core speeds of the non-K 11700 according to this list. It appears the Engineering samples are making it into these lists as if they were actual SKUs :



Napkin math based on CB R20 results :

Edit: Fixed my math

3800Mhz = 4683
4700Mhz = 4924

4683/3800 = 1.232 / mhz
4924/4700 = 1.047/ Mhz

1.232/1.047 = 17.7% faster MT Cinebench R20 per Mhz

+17.7% "IPC" as measured by Cinebench vs Skylake



If that's correct or even close to correct, I think it would be safe to say that AMD launched Zen 3 early with little supply and no major OEM support because they won't be able to sell them next year.

Anyway, was really thinking about getting a 10700 since it's down to $280 at Best Buy now, but after looking at that I do believe I'll wait.

 

[JoeRambo]

R15
Skylake 1
Cypress Cove 1.14
Zen3 1.28 (5800X might be boosting to 4.8Ghz as many reviews show so I used 4.8Ghz)

Really great results for Intel. Cinebench workloads fit ZEN cpus like glove, so if Cypress Cove is within 15% in cinebenches it will have ~10% advantage elsewhere.

By 10% i mean final performance for ~5Ghz RKL versus 4.6Ghz all core workloads for ZEN3 and ~5.2 versus 4.8 in ST workloads

 

[ondma]

Looking at those benchmarks from different sites, it looks like everything from the 11700/11700T/11700K and up is only distinguished by its TDP rating (also related to binning), base/single/multi-core multipliers, and locked/unlocked.

The source article, translated, is also revealing. From that translation :

1 - This might not even be a 10900, they are assuming that.

2 - This is a very low end motherboard, and may not even be a 5-series :

View attachment 36214

3 - The clock speeds actually correspond to an 11700T, a SKU that based on 11700T I would expect to find in high end AIO desktop builds and with mini-PCs from OEMs :

View attachment 36215


4 - This chip is clocked below the expected single and multi-core speeds of the non-K 11700 according to this list. It appears the Engineering samples are making it into these lists as if they were actual SKUs :

View attachment 36216

Napkin math based on CB R20 results :


3800Mhz = 4683 (3.8Ghz all core Rocket Lake 8C/16T)
4700Mhz = 4924 (4.7Ghz all core 9900K score from same source)

4683/3800 = 1.232 / mhz (RKL)
4700/4924 = 0.955/ Mhz (SKL)

1.232/0.955 = 1.290 faster MT Cinebench R20 per Mhz

+29.0% "IPC" as measured by Cinebench vs Skylake

Check my math.

If that's correct or even close to correct, I think it would be safe to say that AMD launched Zen 3 early with little supply and no major OEM support because they won't be able to sell them next year.

Anyway, was really thinking about getting a 10700 since it's down to $280 at Best Buy now, but after looking at that I do believe I'll wait.

If you look at my posts, it is easy to see that I am not one of the many Intel bashers in this thread. However, I think you are trying a bit too hard. There is absolutely no way that there is a 29% increase in IPC from SL to RL, unless some specific step was made to optimize for CB, that wont carry over to all workloads.

I am hopeful for 15 to 18 % increase, instead of the 10% the bashers are projecting, but over 20% is just not going to happen.

 

[shady28]

If you look at my posts, it is easy to see that I am not one of the many Intel bashers in this thread. However, I think you are trying a bit too hard. There is absolutely no way that there is a 29% increase in IPC from SL to RL, unless some specific step was made to optimize for CB, that wont carry over to all workloads.

I am hopeful for 15 to 18 % increase, instead of the 10% the bashers are projecting, but over 20% is just not going to happen.

I fixed the math, It's 17.7%.

And no if you were one of the Intel bashing AMD trolls, you'd be on my ignore list. I think there are 4 or 5 of them that post in this thread regularly.

 

[inf64]

Really great results for Intel. Cinebench workloads fit ZEN cpus like glove, so if Cypress Cove is within 15% in cinebenches it will have ~10% advantage elsewhere.

By 10% i mean final performance for ~5Ghz RKL versus 4.6Ghz all core workloads for ZEN3 and ~5.2 versus 4.8 in ST workloads

I am not that optimistic but we will find out soon . IMO it will be a second tier to Zen3 parts in everything but OCing. Maybe OCed it can push above stock Zen3.

 

[Bam360]

It's true that Cinebench fits Zen arch like a glove, but the delta is not as large for a 15% deficit to be erased in most other workloads. I would say RKL needs to be within 5-7% in Cinebench ST, this would probably mean parity in many other workloads and a win in some like Gaming. Of course I am talking lightly threaded workloads.

 

[shady28]

Looks like CoolerMaster figured out a good way to use all that heat from Comet Lake CPUs.

Yes I like to make fun of Intel sometimes too :

Introducing the KFConsole | Cooler Master

The KFConsole has arrived. Forged from the fires of the KFC ovens and built by Cooler Master from the ground up, there has never been a tastier way to experience the latest titles in stunning 4k, 240fps.

landing.coolermaster.com

 

[Hulk]

I added some additional scores from the Anandtech bench. Keep in mind if Zen 3 finishes this task in 100 seconds then Tiger Lake finished in 95 seconds. Assuming we can trust these clockspeeds, then...
I'm seeing Tiger Lake ~5% behind Zen 3.
Rocket Lake is about 7.5% behind Zen 3.
Ice Lake about 11% behind Zen 3.
Skylake, Kaby Lake, Kaby Lake R, Coffee Lake, Comet Lake and the rest of the Sky derivatives are about 20% behind Zen 3.

Again, assuming these clocks (and scores) are accurate Comet Lake is about 13% faster than Skylake.

Cinebench 20 ST	Core	Clockspeed	Cinebench R20 ST	MHZ/Cinemark	Rel. Performance	Core	% behind Zen 3
5800x (Zen 3)	Zen 3	4800	624	7.69	100.0%	Zen 3
i7-1185G7 @28W Tiger Lake (Willow Cove) - Anandtech Bench	Tiger Lake	4800	593	8.09	95.0%	Tiger Lake	5.0%
Rocket Lake (Cyprus Cove)	Rocket Lake	4400	529	8.32	92.5%	Rocket Lake	7.5%
1065-G7 MS Surface Laptop 3 - Ice Lake - (Sunny Cove) - Anandtech Bench	Ice Lake	3900	449	8.69	88.6%	Ice Lake	11.4%
i3-6300 - Skylake (Original!) - Anandtech Bench	Skylake	3800	399	9.52	80.8%	Skylake	19.2%
Skylake	Skylake	5000	512	9.77	78.8%	Skylake	21.2%
5775C - Broadwell - Anandtech Bench	Broadwell	3700	365	10.14	75.9%	Broadwell	24.1%
4770k - Haswell - Anandtech Bench	Haswell	3900	375	10.40	74.0%	Haswell	26.0%

 

[shady28]

So there was a screenshot of bilibili's Rocket Lake test config that was in Chinese. I couldn't read it of course and google translate doesn't work on images, but I found a way to convert to Chinese and back to english.

The test system was running its RAM at DDR4-2133 for those RKL benchmarks.

Here's the translation :

 

[Bouowmx]

Good catch but memory frequency doesn't affect Cinebench much at all. Don't know about CPU-Z benchmark.

 

[DrMrLordX]

MATLAB is but one software package that uses MKL. There's Anaconda, Mathematica, ANSYS etc. The situation with MKL performance on Zen is fluid. Some things work quite well while others do not and it changes all the time. As long as this is the case Intel is a viable option for workstation use. Not all use-cases require 16 cores of the 5950X.

I think we can both agree that the "fluid" situation isn't a hardware problem? Hard to really give credit to Intel hardware for that mess.

Couldn't care less what marketing does. What the engineering teams could have done is come up with something better than Skylake+++. For those who believe Intel halted CPU development due to 10nm delays, that is wrong, they kept going but all they could come up with was more Skylake.

In theory, yes, in practice, no. None of us will really know how many resources they kept throwing at 14nm CPU development, but I have a feeling they might have been crazy enough to put all their teams on designs targeted at 10nm.

If that's correct or even close to correct, I think it would be safe to say that AMD launched Zen 3 early with little supply and no major OEM support because they won't be able to sell them next year.

The launch was late, not early. Also you're talking about an 8c chip with the MT performance of a 10900k. That won't affect sales of any competitors' CPUs much.

 

[tamz_msc]

I think we can both agree that the "fluid" situation isn't a hardware problem? Hard to really give credit to Intel hardware for that mess.

It's not a hardware problem but a problem of software support in performance-critical applications. With the situation like this there is no merit to your claim that nobody in their right mind would use Rocket Lake over Zen 3 in workstation applications.

 

[DrMrLordX]

It's not a hardware problem but a problem of software support in performance-critical applications. With the situation like this there is no merit to your claim that nobody in their right mind would use Rocket Lake over Zen 3 in workstation applications.

. . . until the people who actually use said applications realize they have better hardware available to them that is being artifically walled-off by Intel constantly breaking their MKL, which puts pressure on the application developers to fix the problem, which has already happened more than once with MATLab.

 

[tamz_msc]

. . . until the people who actually use said applications realize they have better hardware available to them that is being artifically walled-off by Intel constantly breaking their MKL, which puts pressure on the application developers to fix the problem, which has already happened more than once with MATLab.

Intel software running better on Intel hardware isn't a problem. AMD not providing competing libraries that deliver the same capability that makes their hardware inferior to Intel is the real problem.