Question Alder Lake - Official Thread

[Hulk]

With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.

 

[uzzi38]

So computerbase measures the power draw in a way that they include the uncore in the numbers?
Or are they just typing in a limit in the mobo? Is that subtracting or adding the uncore load?

They just set PL1==PL2==88W for Intel and PPT to 88W for AMD effectively. Both power limits are full package power.

 

[Heartbreaker]

Link please? I can see where they overclocked in their reviews of each, I can't see any measurements of power draw under that OC.

The 4.6GHz all core OC shows up in every slide including power:

https://tpucdn.com/review/amd-ryzen-7-5800x/images/power-multithread.png

https://tpucdn.com/review/amd-ryzen-7-5700g/images/power-multithread.png

At 4.6GHz all core, 5800x is using 178 W, 5700G is using 181 W. This is despite the 5700G having worse performance.

At 4.6GHz for both, the 5700G uses slightly more power, while have slightly worse performance.

 

[Asterox]

View attachment 52515

Very intresting or informative i5 12400F review.French website has blah desing, but it is ok if the information is grouped and accurately highlighted.

Test • Intel Core i9-12900K/12900F / i7-12700K / i5-12600K/12600/12400F & Z690/B660

Test des processeurs Intel Core i9-12900K, i9-12900F, i7-12700K, i5-12600K, i5-12600 et i5-11400F ainsi que leur chipset compagnon Z690, B660 et nouveau socket LGA 1700. Comparaison avec les précédentes générations de processeurs Intel et AMD sur des jeux, encodage, traitement d'image et vidéo...

www.comptoir-hardware.com

R5 and i5 was cooled by Noctua CPU air cooler.

Ryzen 5 5600X/DDR4 memory, is faster %

- Veracrypt, 11%
- Stockfish, 23%
- After Effects, 10%
- Blender, 4%
- Vegas Pro, 17%
- LightRoom, 3%
- Handbrake H264, 10%
- Handbrake H265, 9%
- Cinema 4D R23, 1%
- Visual Studio 2019/Compilation Unreal Engine, 5%
- Compilation Linux, 9%
- CPU-Z Multithread same score

i2 12400F/DDR5 memory, is faster %

- p7Zip, 25%
- Tensorflow, 20%
- DxO Photolab, 3%
- Arnold, 4%
- Compilation GCC, 9%
- CPU-Z Singlethread, 6%
- Cinebench R23 Singlethread, 11%
- Cinebench R23 Multithread, 2%
- CPU-Z Multithread same score

On average(page 18-20), R5 vs i5 gaming performanse is very similar.In game FPS advantage depends a lot on the particular game used.

 

[mikk]

Very intresting or informative i5 12400F review.French website has blah desing, but it is ok if the information is grouped and accurately highlighted.

Test • Intel Core i9-12900K/12900F / i7-12700K / i5-12600K/12600/12400F & Z690/B660

Test des processeurs Intel Core i9-12900K, i9-12900F, i7-12700K, i5-12600K, i5-12600 et i5-11400F ainsi que leur chipset compagnon Z690, B660 et nouveau socket LGA 1700. Comparaison avec les précédentes générations de processeurs Intel et AMD sur des jeux, encodage, traitement d'image et vidéo...

www.comptoir-hardware.com

R5 and i5 was cooled by Noctua CPU air cooler.

Ryzen 5 5600X/DDR4 memory, is faster

After 56 seconds this 12400F drops down to 3.4 Ghz and that's why all the longer benchmarks in particular are not that good, I think there is something wrong be it a bios problem or 12400F ES related.

 

[Abwx]

Very intresting or informative i5 12400F review.French website has blah desing, but it is ok if the information is grouped and accurately highlighted.

Test • Intel Core i9-12900K/12900F / i7-12700K / i5-12600K/12600/12400F & Z690/B660

Test des processeurs Intel Core i9-12900K, i9-12900F, i7-12700K, i5-12600K, i5-12600 et i5-11400F ainsi que leur chipset compagnon Z690, B660 et nouveau socket LGA 1700. Comparaison avec les précédentes générations de processeurs Intel et AMD sur des jeux, encodage, traitement d'image et vidéo...

www.comptoir-hardware.com

R5 and i5 was cooled by Noctua CPU air cooler.

Ryzen 5 5600X/DDR4 memory, is faster

- Veracrypt
- Stockfish
- After Effects
- Blender
- Vegas Pro
- LightRoom
- Handbrake H264
- Handbrake H265
- Cinema 4D R23
- Visual Studio 2019/Compilation Unreal Engine 4.22
- Compilation Linux
- CPU-Z Multithread little faster

i5 12400F/DDR5 memory, is faster

- p7Zip
- Tensorflow
- DxO Photolab
- Arnold
- Compilation GCC
- CPU-Z Singlethread
- Cinebench R23 Singlethread/Multithread little faster

On average(page 18-20), R5 vs i5 gaming performanse is very similar.In game FPS advantage depends a lot on the particular game used.

View attachment 52561

View attachment 52562

View attachment 52563

View attachment 52564

View attachment 52565

So on Cinebench R23 the 5600X is slower but in an actual application of Cinema 4D R23 it is faster....Interesting to say the least.

 

[LightningZ71]

The 4.6GHz all core OC shows up in every slide including power:

https://tpucdn.com/review/amd-ryzen-7-5800x/images/power-multithread.png

https://tpucdn.com/review/amd-ryzen-7-5700g/images/power-multithread.png

At 4.6GHz all core, 5800x is using 178 W, 5700G is using 181 W. This is despite the 5700G having worse performance.

At 4.6GHz for both, the 5700G uses slightly more power, while have slightly worse performance.

Hmm, if I had a guess, I would say that the smaller L3 cache of the 5700g means that it had to go to main memory a lot more often that the 5800x. Memory access introduces latencies and performance hits. It also is more expensive in power draw. Both products are done on the same process, save for the IOD. I'm not surprised that the 5700g performed worse while consuming more power.

 

[mikk]

So on Cinebench R23 the 5600X is slower but in an actual application of Cinema 4D R23 it is faster....Interesting to say the least.

Cinebench R23 isn't really affected from the drop off after 56 seconds, it's much shorter. It's logical that CB R23 runs much better versus the longer benchmarks.

 

[jpiniero]

After 56 seconds this 12400F drops down to 3.4 Ghz and that's why all the longer benchmarks in particular are not that good, I think there is something wrong be it a bios problem or 12400F ES related.

If the base clock is 3.4, there's nothing unusual going on.

 

[Heartbreaker]

So on Cinebench R23 the 5600X is slower but in an actual application of Cinema 4D R23 it is faster....Interesting to say the least.

I wouldn't read much into that. They are effectively the same. The Margins of victory back and forth are 2% and 3%.

 

[TheELF]

Cinebench R23 isn't really affected from the drop off after 56 seconds, it's much shorter. It's logical that CB R23 runs much better versus the longer benchmarks.

They are showing 241W power draw for the 12900k which means that they are running at all the turbo all the time mode.
it would be pretty weird if they had the other CPUs in a different mode.
Unless they used the exact same scene in cinema as the benchmark uses there are bound to be differences, we know that from the various blender benches where different scenes can have quite different scores.

 

[Asterox]

After 56 seconds this 12400F drops down to 3.4 Ghz and that's why all the longer benchmarks in particular are not that good, I think there is something wrong be it a bios problem or 12400F ES related.

Hm i doubt it, keep in mind All Core Turbo CPU frequencie diference.This is not high clocked i5 from K CPU series.

- i5, up to 4ghz All core Turbo
4.4ghz Singlecore
- R5, 4.4-4.5ghz All Core turbo
4.6ghz Singlecore

Tests are in seconds. So what, in Arnold test faster i5 is 4ghz All Cores? But in Handbrake or Blender, hm it is only 3.4ghz All Core Turbo.

I think you expect too much from the Golden Cove 6/12 4ghz CPU All core turbo.

 

[uzzi38]

The 4.6GHz all core OC shows up in every slide including power:

https://tpucdn.com/review/amd-ryzen-7-5800x/images/power-multithread.png

https://tpucdn.com/review/amd-ryzen-7-5700g/images/power-multithread.png

At 4.6GHz all core, 5800x is using 178 W, 5700G is using 181 W. This is despite the 5700G having worse performance.

At 4.6GHz for both, the 5700G uses slightly more power, while have slightly worse performance.

Huh, that's surprising.

Not that it changes much, if anything it just shows how Cezanne's freq/voltage scaling is worse than Vermeer's at higher clocks. After all, the Vermeer chip there is also set to a higher voltage - 1.35v vs 1.4v yet pulls essentially the same power, which just simply doesn't make sense.

I already provided a pretty clear perf/W chart with Renoir vs Matisse, and to show that Cezanne is more like Renoir than it is Matisse/Vermeer, here:s the same graph for the 5700G:

https://media.discordapp.net/attachments/697380330864115792/877430531132899338/E8b6BM1X0AUjF7z.png

 

[Geegeeoh]

After 56 seconds this 12400F drops down to 3.4 Ghz and that's why all the longer benchmarks in particular are not that good, I think there is something wrong be it a bios problem or 12400F ES related.

IIRC only K CPU have infinite turbo, "vanilla" CPU will go back to base clock after the burst.

 

[jpiniero]

IIRC only K CPU have infinite turbo, "vanilla" CPU will go back to base clock after the burst.

You can still set the locked parts to have infinite Tau, just that the recommended stock setting is to not do that.

 

[TheELF]

You can still set the locked parts to have infinite Tau, just that the recommended stock setting is to not do that.

That's true for the K parts as well.

Intel® Core™ Ultra and Intel® Core™ Processors Technical Resources

Intel® Core™ Ultra and Intel® Core™ Processors technical resources list includes applications notes, datasheets, packing information, product briefs, and more.

www.intel.com

volume 1 page 76

 

[DrMrLordX]

1: Apple M1 core cluster at 30 watts is the highest power it will ever draw, therefore the lowest efficiency it will operate at. 30W for a 12900k is about as low as it will go, and around the highest efficiency it will run at. You just attempted to make an efficiency claim on the most misleading single data point available.

Then maybe joules-per-job should be used at better power efficiency points then?

So on Cinebench R23 the 5600X is slower but in an actual application of Cinema 4D R23 it is faster....Interesting to say the least.

Cinebench benchmarks often use a slightly different codebase than actual shipping versions of Cinema 4D. That was definitely true in the R15 days.

Cinebench R23 isn't really affected from the drop off after 56 seconds, it's much shorter. It's logical that CB R23 runs much better versus the longer benchmarks.

R23 loops for 10 minutes by default unless you defeat that behavior. It was intended to generate a score that would resist the effects of short-tau turbo periods.

 

[Tarkin77]

That is the highest single thread score I've seen on a 5800x.

My CO is more tuned for higher multi and SMT is on.

View attachment 52501

System manufacturer System Product Name - Geekbench

Benchmark results for a System manufacturer System Product Name with an AMD Ryzen 7 5800X processor.

browser.geekbench.com

My 5800X can do over 1800 single and 12000 multi

Micro-Star International Co., Ltd. MS-7B79 - Geekbench

Benchmark results for a Micro-Star International Co., Ltd. MS-7B79 with an AMD Ryzen 7 5800X processor.

browser.geekbench.com

 

[Asterox]

I wouldn't read much into that. They are effectively the same. The Margins of victory back and forth are 2% and 3%.

I am not gamer, so i updated only original listed aplication scores with % numbers.

Page 28 - Question - Alder Lake - Official Thread

Page 28 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[mikk]

IIRC only K CPU have infinite turbo, "vanilla" CPU will go back to base clock after the burst.

I know. The point is it shouldn't drop down to 6x3.4 Ghz to stay within 65W PL1. This isn't static, some apps are more heavy than others.

 

[jpiniero]

I know. The point is it shouldn't drop down to 6x3.4 Ghz to stay within 65W PL1. This isn't static, some apps are more heavy than others.

The cheapest i5 (which is the 12400F I guess in this case) could be binned that loosely for that to happen IMO.

 

[dmens]

Then maybe joules-per-job should be used at better power efficiency points then?

Unless your goal is lowest energy at any performance, you need to anchor joules expended with a performance target.

 

[mikk]

The cheapest i5 (which is the 12400F I guess in this case) could be binned that loosely for that to happen IMO.

I don't think so. We have seen a 12400 leak running at 4.0 Ghz in AIDA64 FPU using 78W, it would probably need 70W in Cinebench MT and roughly 3.9 Ghz to stay within 65W. From 4.0 or 3.9 Ghz down to 3.4 Ghz is a different world. If it drops down to 3.4 Ghz in every test after 56s something is not right which isn't even a problem, it's an ES 12400 and the bios possibly isn't fully optimized for the upcoming 65W SKUs.

 

[DrMrLordX]

Unless your goal is lowest energy at any performance, you need to anchor joules expended with a performance target.

When comparing the M1 Max and 12900k, it's difficult to set them to a common performance target without (as you observed) skewing efficiency results. You can't really get an M1 Max to burn 200W+ on CBR23, but you sure can do it with a 12900k!

 

[jpiniero]

I don't think so. We have seen a 12400 leak running at 4.0 Ghz in AIDA64 FPU using 78W, it would probably need 70W in Cinebench MT and roughly 3.9 Ghz to stay within 65W. From 4.0 or 3.9 Ghz down to 3.4 Ghz is a different world. If it drops down to 3.4 Ghz in every test after 56s something is not right which isn't even a problem, it's an ES 12400 and the bios possibly isn't fully optimized for the upcoming 65W SKUs.

If the base clock is set at 3.4 it was set for a reason. It's possible that the non-F is better binned than the F.

 

[dmens]

When comparing the M1 Max and 12900k, it's difficult to set them to a common performance target without (as you observed) skewing efficiency results. You can't really get an M1 Max to burn 200W+ on CBR23, but you sure can do it with a 12900k!

Because that is the wrong way to look at CB MT. CB MT is an exercise in perf with core scaling, that is fundamentally a question of how many cores you can cram into a socket of some number of tolerable watts. If you set an anchor as some perf number but compare two different die sizes, it doesn't say a whole lot about efficiency. The smaller die will typically lose badly because it is forced to run up the curve to hit that perf number.