Question Alder Lake - Official Thread

Page 117 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

igor_kavinski

Diamond Member
Jul 27, 2020
4,866
3,061
106
Well I'm not a Linux user, but I know for sure that under Windows picture differs a lot to that (this is the prior benchmark version though, I'll run later the current version but I highly doubt things change):




On average, 5950X is 18% faster (consuming clearly less power than 12900K)
Waiting to see your new version benchmark run :)

Also, are you using DDR4 or DDR5 with 12900K? What speed and timings?

If you have a spare SSD, maybe you could run these benchmarks on Clear Linux?
 

MarkPost

Member
Mar 1, 2017
191
165
116
Waiting to see your new version benchmark run :)

Also, are you using DDR4 or DDR5 with 12900K? What speed and timings?

If you have a spare SSD, maybe you could run these benchmarks on Clear Linux?
Same speed/timmings (DDR4) for both systems. Anyways, AFAIK memory type/freq/timmings are basically irrelevant in this type of tasks

About running it under Linux, well in my life used Linux so I know nothing about it, sorry
 

igor_kavinski

Diamond Member
Jul 27, 2020
4,866
3,061
106
Same speed/timmings (DDR4) for both systems. Anyways, AFAIK memory type/freq/timmings are basically irrelevant in this type of tasks
You are gimping your 12900K. For highly multi-threaded tasks, it does best with DDR5-6400 or DDR5-6600 with CL34 or less. Those Phoronix Blender 3.3 benchmarks done with DDR5 of unknown speed.
 

igor_kavinski

Diamond Member
Jul 27, 2020
4,866
3,061
106
About running it under Linux, well in my life used Linux so I know nothing about it, sorry
It's never too late to learn :p

If you have a spare SSD, just download ISO, create bootable USB from ISO, boot and install, login, launch Firefox and download Blender for Linux. Not terribly hard, trust me.
 

igor_kavinski

Diamond Member
Jul 27, 2020
4,866
3,061
106
1663687600724.png
1663687624962.png

Under Windows, 12900K is roughly 7% behind 5950X.

Larabel is taking advantage of patched Linux with better scheduling support for Alder Lake's hybrid core scheme. Windows scheduling issues may be preventing ADL from performing as well as under Linux.
 

Abwx

Diamond Member
Apr 2, 2011
9,755
2,096
136
View attachment 67823
View attachment 67825

Under Windows, 12900K is roughly 7% behind 5950X.

Larabel is taking advantage of patched Linux with better scheduling support for Alder Lake's hybrid core scheme. Windows scheduling issues may be preventing ADL from performing as well as under Linux.
Larabel use short duration Blender scenes such that the 12900K perf doesnt tank too much over time, at Computerbase the 5950X is 16% faster in their Blender quick bench whose duration is about 8mn.

 

Exist50

Golden Member
Aug 18, 2016
1,162
1,158
136
Larabel use short duration Blender scenes such that the 12900K perf doesnt tank too much over time, at Computerbase the 5950X is 16% faster in their Blender quick bench whose duration is about 8mn.

If PL1==PL2, performance shouldn't change with load duration.
 

Abwx

Diamond Member
Apr 2, 2011
9,755
2,096
136
Last edited:

igor_kavinski

Diamond Member
Jul 27, 2020
4,866
3,061
106
It shouldnt in a world of AIO, here the temp reach 100°C after 150s...
HolythrottleCow!

Intel needs to find a way to convert all that extra thermal energy into something useful. Then all of a sudden, people will be trying to force their Ryzens to 100 degrees too! :D

Maybe a miniature thermoelectric generator connected to the CPU heatsink to reduce the bill?
 

Exist50

Golden Member
Aug 18, 2016
1,162
1,158
136
It shouldnt in a world of AIO, here the temp reach 100°C after 150s...

The Blender temp graph is on the meddle of the page :

That's for unlimited power... If you're going to do that, yeah, you'll probably have an AIO. And why mix data between the two?
 

Abwx

Diamond Member
Apr 2, 2011
9,755
2,096
136
HolythrottleCow!
Technically as long as it s above base frequency it s not throttling, just peak frequency not being sustained.

That being said some members here say that 250W is a limit for a good AIO, no wonder that such powers cant be sustained for long on air coolers.

That's for unlimited power... If you're going to do that, yeah, you'll probably have an AIO. And why mix data between the two?
That s not for unlimited power, that s at stock settings, read above my remark about thermal dissipation of cooling apparatuses..

One has to aknowledge that it s a factory hugely overclocked CPU, all benches where performance doesnt tank are made with AIOs, on air it would be the same with enough CB R23 runs to reach the temp limit, a single run or two take advantage of the thermal inertia that last 150s with the cooler of Computerbase.
 
Last edited:

igor_kavinski

Diamond Member
Jul 27, 2020
4,866
3,061
106
I just had an idea. Why aren't AIO tubes lined with an aluminum or even copper netting? That way the coolant can lose some heat on its way to the radiator.
 

MarkPost

Member
Mar 1, 2017
191
165
116
ok I run the latest version. Both systems with 24/7 OC configs.



Total Score for my oced 12900KF is 418.87



Total Score for my oced 5950X is 475.75

So 5950X is about 14% faster

As a bonus, CPU package power consumption while running the benchmark:




perf/power ratio is a no contest between both
 

Abwx

Diamond Member
Apr 2, 2011
9,755
2,096
136
Those thermal numbers were literally from the no limits 8+8 config...
That change nothing, with PL1 = PL2 = 241W the CPU stay at 241W, without limit it goes up to 280-290W, you think that 40-50W more make any other difference than reducing a little the time it take to hit the critical temp..?.

Their blender test lats about 8mn, that s way more than 150s, assuming that it s on unlimited mode then at 241W it would require about 200-220s to hit the critical temp.
 

igor_kavinski

Diamond Member
Jul 27, 2020
4,866
3,061
106
ok I run the latest version. Both systems with 24/7 OC configs.



Total Score for my oced 12900KF is 418.87



Total Score for my oced 5950X is 475.75

So 5950X is about 14% faster

As a bonus, CPU package power consumption while running the benchmark:




perf/power ratio is a no contest between both
May I request one more battle? Both Zen 3 and ADL at 8C/16T.
 

Exist50

Golden Member
Aug 18, 2016
1,162
1,158
136
That change nothing, with PL1 = PL2 = 241W the CPU stay at 241W, without limit it goes up to 280-290W, you think that 40-50W more make any other difference than reducing a little the time it take to hit the critical temp..?.
Uh, yes, another 20% more power absolutely matters. And the performance numbers you're quoting are from 125W/241W, Iirc. Why mix the two configs only to highlight the worst half of each?
 

Abwx

Diamond Member
Apr 2, 2011
9,755
2,096
136
Uh, yes, another 20% more power absolutely matters. And the performance numbers you're quoting are from 125W/241W, Iirc. Why mix the two configs only to highlight the worst half of each?
Seems that you are finding all reasons to balance what is actually an erroneous perception of the 12900K actual perf.

Without limit for the 12900K the 5950X is still 15% faster in Blender.

If set at 241W the 5950X is 16% faster.

If set as you say at 125/241W then the 5950X is 30% faster.

Click on Blender and then on "+88 Eintrage" :

 
  • Like
Reactions: lightmanek

Exist50

Golden Member
Aug 18, 2016
1,162
1,158
136
Seems that you are finding all reasons to balance what is actually an erroneous perception of the 12900K actual perf.

Without limit for the 12900K the 5950X is still 15% faster in Blender.

If set at 241W the 5950X is 16% faster.

If set as you say at 125/241W then the 5950X is 30% faster.

Click on Blender and then on "+88 Eintrage" :

Thanks. On mobile, so I was having trouble browsing their site. Still not sure I see what point you're trying to make about difficulty cooling a chip that's essentially being overclocked.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
23,407
12,379
136
Thanks. On mobile, so I was having trouble browsing their site. Still not sure I see what point you're trying to make about difficulty cooling a chip that's essentially being overclocked.
I don't see anything about cooling. I see that he is saying regardless of the power settings, the 5950x beats the 12900k, just depends on what power settings as to what % it wins by. In the example he is using.
 

Exist50

Golden Member
Aug 18, 2016
1,162
1,158
136
But you were quoting 2919, thats what I was responding to.

Edit: I am using a 27 inch monitor. You are using mobile. I can see how its a little confusing.
Ah, that response to 2919 was a thanks for clarifying the information. Then I tacked on the callback to 2909 without a quote.
 

ASK THE COMMUNITY