Discussion Intel current and future Lakes & Rapids thread

[Markfw]

28K score requires an overclock on the 5950X (and all-core clock speeds around 4.4 GHz). At stock, its score is in the 25K range

CPU Benchmarks and Hierarchy 2025: CPU Rankings

We've run thousands of CPU benchmarks on all new and older Intel and AMD CPUs and ranked them.

www.anandtech.com

Well, that may be normal, but the site does not say anywhere that the run was overclocked. Its possible it was stock. Where did you see it was OC'ed ??

And I still don't believe anything that a random site says, with NO specs of the details of the run. For all we know, its all photoshop.

 

[Accord99]

Well, that may be normal, but the site does not say anywhere that the run was overclocked. Its possible it was stock. Where did you see 4.4 ??

I have a 5950x too, I get a score of about 28.6K using PBO2 overclocking (by increase the package power limit to 250W) and a -9 offset in the curve optimizer which results in all-core clock speeds of 4.45GHz when running CB23:

 

[jpiniero]

I'm sorry. I was a little late. It's just 0+8 is reserved. It doesn't indicate a SKU exists.

It makes sense to me even if they sell it off label or use Atom Embedded brand. There's going to be enough all P core busted dies.

 

[Markfw]

I have a 5950x too, I get a score of about 28.6K using PBO2 overclocking (by increase the package power limit to 250W) and a -9 offset in the curve optimizer which results in all-core clock speeds of 4.45GHz when running CB23:

View attachment 52509

But what memory speed are you using ? I have 4000 cl16. I makes quite a difference. I see where you are coming from, but I still don't trust that chinese sites numbers.

 

[JasonLD]

And I still don't believe anything that a random site says, with NO specs of the details of the run. For all we know, its all photoshop.

He seemed to pretty clearly explain what components were used for testing. (Asus Z690 RoG Hero with Trident Z DDR5 6000/c40 and Asus Z690 TUF with Trident Royal DDR4 3600/c16 and RTX3090 FE for GPU).

 

[Abwx]

Also impressive is the 25244 score with package power of 117W, which is better than the 5950x.

Given the high performance/watt of Alder Lake at reasonable clock speeds, we may see AMD "recommend" 200+W power settings in reviews for its 16-core CPUs going forward.

By undervolting as much as possible he did eat on the stability margin to get those results, so they are meaningless, lower power but not stable at all on the long run.

 

[insertcarehere]

By undervolting as much as possible he did eat on the stability margin to get those results, so they are meaningless, lower power but not stable at all on the long run.

FWIW, the reviewer stated that it's stable with AIDA 64 FPU at those UVs but your point is valid.

The reviewer also tested putting the wimpy Intel stock LGA1200 cooler on the 12900k.


No manual UV was done, package power stabilized at ~140w, with a Cinebench R23 score of over 25k.

Reviewer here added more nuance in the review than most by showing perf in power-limited scenarios rather than just further reinforcing the "OMG 12900k is a fast furnace you need a watercooler in order to run it!!!".

 

[Abwx]

FWIW, the reviewer stated that it's stable with AIDA 64 FPU at those UVs but your point is valid.

The reviewer also tested putting the wimpy Intel stock LGA1200 cooler on the 12900k.

View attachment 52524
No manual UV was done, package power stabilized at ~140w, with a Cinebench R23 score of over 25k.

If it was set at 140W with default settings then HVinfo wouldnt display 4096W PL1, wich is diplayed when all brakes are removed for manual settings, it would display PL1 = PL2 = 140W or whatever the available factory power limits from Intel.

 

[insertcarehere]

If it was set at 140W with default settings then HVinfo wouldnt display 4096W PL1, wich is diplayed when all brakes are removed for manual settings, it would display PL1 = PL2 = 140W or whatever the available factory power limits from Intel.

My point being that the reviewer (probably) plopped a stock Intel cooler on the 12900k w/ zero changes to default MB settings (hence 4096w PL1), and this was the result given the thermal limits of the Intel cooler.

 

[JoeRambo]

I have a 5950x too, I get a score of about 28.6K using PBO2 overclocking (by increase the package power limit to 250W) and a -9 offset in the curve optimizer which results in all-core clock speeds of 4.45GHz when running CB23:

View attachment 52509

It's possible to do better with custom undervolt and fixed voltage for 4.4Ghz:




~145W for 28.8k score.

Another point:



~125W for 27.2k score

And another point, matching that custom undervolted 12900K:



~105W for 25.6k score, 3.9Ghz

 

[Abwx]

My point being that the reviewer (probably) plopped a stock Intel cooler on the 12900k w/ zero changes to default MB settings (hence 4096w PL1), and this was the result given the thermal limits of the Intel cooler.

He did a test at manual settings and fixed frequency, you can see that the single core score is significantly lower than what is measured elsewhere, 1750 instead of 2025.

 

[deasd]

MSI official document, E-core enable/disable test using AIDA FPU:

 

[pakotlar]

Also impressive is the 25244 score with package power of 117W, which is better than the 5950x.

Given the high performance/watt of Alder Lake at reasonable clock speeds, we may see AMD "recommend" 200+W power settings in reviews for its 16-core CPUs going forward.

Yeah they’ll also need to recommend higher TDC (thermal design current). It takes me about 120-130 TDC to hit 27726, and that is about 180W, so a bit better than what is seen above with Alder Lake.
I think Alder Lake is going to be very exciting in the laptop space, seems to scale down with voltage very well. For desktops, it appears slightly less exciting at the high end, and really nice for the midrange parts. Actually 12900K undervolted seems exciting.

 

[Accord99]

It's possible to do better with custom undervolt and fixed voltage for 4.4Ghz:

I think the next few weeks and months will be interesting in terms of under-volting as some users and reviewers try to optimize Alder Lake for performance/watt and see how that compares with the best under-volts for the 5900X and 5950X.

 

[Abwx]

Also impressive is the 25244 score with package power of 117W, which is better than the 5950x.

At stock a 5950X does 26100 pts and consume 125W to do so.

Intel Core i9-12900K, i7-12700K & i5-12600K im Test: Benchmarks in Anwendungen

Intel Alder Lake im Test: Benchmarks in Anwendungen / Ein rekordverdächtiger Testumfang / Windows 10 vs. Windows 11

www.computerbase.de

 

[JoeRambo]

I think Alder Lake is going to be very exciting in the laptop space, seems to scale down with voltage very well. For desktops, it appears slightly less exciting at the high end, and really nice for the midrange parts. Actually 12900K undervolted seems exciting.

On desktop undervolt dips down into power twice. Not only you reduce the power for P cores needed to reach set frequency, the excess power wasted due to E cores being on same voltage plane is also reduced. The "optimum" point is should be nearer to what E cores require for ~4Ghz.

10ESF looking very nice actually, this repurposed mobile CPU is doing better on desktop than most expected.

 

[pakotlar]

On desktop undervolt dips down into power twice. Not only you reduce the power for P cores needed to reach set frequency, the excess power wasted due to E cores being on same voltage plane is also reduced. The "optimum" point is should be nearer to what E cores require for ~4Ghz.

10ESF looking very nice actually, this repurposed mobile CPU is doing better on desktop than most expected.

Good point, wonder why they don’t have 2 different power planes for p and e cores

 

[mikk]

At stock a 5950X does 26100 pts and consume 125W to do so.

Intel Core i9-12900K, i7-12700K & i5-12600K im Test: Benchmarks in Anwendungen

Intel Alder Lake im Test: Benchmarks in Anwendungen / Ein rekordverdächtiger Testumfang / Windows 10 vs. Windows 11

www.computerbase.de

They say 142 watts.

 

[uzzi38]

They say 142 watts.

No 5950X out there will actually pull 142W in Cinebench. TDC limits are too restrictive, so the cores sit around 3.7-3.8GHz at 120-130W.

 

[Abwx]

They say 142 watts.

That s 125W, the delta from iddle to load is 143W, account the PSU and VRMs converters losses (x 0,77) and add 15W for the CPU iddle comsumption, the 5950X consume less than the 5900X and the same as a 5800X.

Btw, they use Cinebench R20 for power tests unless specified otherwise.

Likewise the 12900K used 226W in their CB run, assuming CPU iddle power is 5W.

Intel Core i9-12900K, i7-12700K & i5-12600K im Test: Leistungsaufnahme, Temperaturen und OC

Intel Alder Lake im Test: Leistungsaufnahme, Temperaturen und OC / Leistungsaufnahme in Spielen / Leistungsaufnahme in Anwendungen

www.computerbase.de

No 5950X out there will actually pull 142W in Cinebench. TDC limits are too restrictive, so the cores sit around 3.7-3.8GHz at 120-130W.

IIRC it hoover around 4GHz.

 

[eek2121]

So, with Intel apparently planning to double the mont cores in Raptor Lake, do we think Raptor Cove will be more power efficient? IIRC The latest leaks had PL2 listed as 253W, which is 12W higher than ADL-S, and they appear to still be using Gracemont for the small cores.

Curiously, PL4 for RPL-S is supposed to be 45W lower than ADL-S…

 

[Mopetar]

Probably. Intel is already capable of being efficient, but they wanted impressive numbers so pulled out the stops and pushed the i9 to the limits. Only this time it's worth that power use and if you look at the gaming results it doesn't even need all of that power.

If they would have reigned in the power they still could have beat AMD, but it wouldn't be by double digits in a lot of the places they win and that's probably a worse headline than the huge power draw that they chips can pull down. They already have that kind of reputation right now anyways and people buying Intel don't seem to care too much anyways.

 

[Zucker2k]

Complementary power delivery system for RPL?

https://twitter.com/x/status/1428233148870111233

 

[dullard]

Probably. Intel is already capable of being efficient, but they wanted impressive numbers so pulled out the stops and pushed the i9 to the limits. Only this time it's worth that power use and if you look at the gaming results it doesn't even need all of that power.

I've never quite figured out the logic of buying an overclockable chip and then complaining when it uses more power. If too much power is your concern, lower the PL2 value or wait for the non-overclockable chips to be launched next quarter.

Rocket Lake cores might be a hair more efficient, but I don't think that is really something to count on. ~10% more speed on the performance cores seems to be Intel's goal (combination of slightly higher turbos and IPC improvements). Only if you do use heavily-threaded applications, will Raptor Lake be able to do more work with more cores in about the same amount of power.

 

[epsilon84]

Probably. Intel is already capable of being efficient, but they wanted impressive numbers so pulled out the stops and pushed the i9 to the limits. Only this time it's worth that power use and if you look at the gaming results it doesn't even need all of that power.

If they would have reigned in the power they still could have beat AMD, but it wouldn't be by double digits in a lot of the places they win and that's probably a worse headline than the huge power draw that they chips can pull down. They already have that kind of reputation right now anyways and people buying Intel don't seem to care too much anyways.

Basically this. Even at 14nm, Intel was fairly efficient if not for the near 5GHz boost clocks and the exponential climb in power that came with that.

Optimal performance/watt for Intel 10nm (sorry, Intel 7 ) is still around the ~4GHz mark it seems. Run a 12900K at 4.0 on the P cores (with an appropriate undervolt) and it would be comparable or better in performance/watt to a 5950X, but it would lose a lot more MT benchmarks than it wins. It would still beat a 5900X, but no one ever remembers the silver medalist!

I'm sure it was a calculated decision to cop the flak for the high power draw to have the 'fastest desktop CPU in the world'. That's the headline grabbing feat 'Intel defeats AMD' not 'it takes 250W to do so'