Arrow Lake Retested with Latest 24H2 Updates and 0x114 Microcode

[LittleBro]

1) Idle power draw - the 9950x draws 38% more power

Unfortunately, this is the problem that concerns all chiplet-based Zen 4/5 CPUs. AMD should really improve at this point.
You obviously don't buy workstation CPU in order to leave it at idle, but okay, valid point there.

2) ST efficiency - the 285k is 36% more efficient. That's absolutely insane btw.

ST efficiency in that review is based on one test only (CineBench 2024 1T).
Unfortunately, there is no info on power draw during CB test. I found it in graph that the consumption is +/- the same as with MP3 Encode.

In CineBench 2024 1T:
285K, 26W, 145 points = 5.58 pts/W
9950X, 36W, 143 points = 3.97 pts/W
285K is more efficient in CB 2024 1T by 28%

In review, power consumption in single thread workload is measured in MP3 Encode:
9950X, 58.8s x 36w = 2116.8 J
285K, 66.9s x 26W = 1739.4 J
285K is about 17.2% more efficient than 9950X in MP3 Encode.
So, it depends ...

285K is more efficient than 9950X based on those two calculations above. Valid point there.

3) MT efficiency - they are on par, 0.0008% difference between them.

Both 285K and 9950X were designed mainly for multithreaded workloads, as their core counts say.
285K was supposed to be beating 9950X thanks to 8 more cores while also being more efficient thanks to node advantage.
That's was I was trying to point out. No valid point there.

4) Gaming efficiency - the 285k is 12% more efficient.

720p:
9950X, 104W, 198.6 fps = 1.91 fps/W
285K, 94W, 192.8 fps = 2,05 fps/W
285K is 7.3% more efficient in 720p gaming based on 14 games tested in that review.
Other games = other results. So, it really depends ... But I'll give you another valid point there.

Based on mentioned review, 285K is more efficient than 9950X in: single thread, gaming, idle. In multithread workload not so much, which is a pitty because that's the main area where both CPUs are focusing on (you don't need more than 8 cores for gaming and/or single thread workloads, right?). Still, it would be worth a while to do a re-bench with AMD's updates for Zen 4/5 at place, since in some areas they represent as much as 5-10% performance increase. Another link.

 

[LittleBro]

Arrow Lake's had three months of Windows and BIOS updates to fix its performance, and my testing shows in some games, it's worse

I thought fixes were supposed to improve things.

www.pcgamer.com

 

[Deleted member 235585]

Unfortunately, this is the problem that concerns all chiplet-based Zen 4/5 CPUs. AMD should really improve at this point.
You obviously don't buy workstation CPU in order to leave it at idle, but okay, valid point there.


ST efficiency in that review is based on one test only (CineBench 2024 1T).
Unfortunately, there is no info on power draw during CB test. I found it in graph that the consumption is +/- the same as with MP3 Encode.

View attachment 380004

In CineBench 2024 1T:
285K, 26W, 145 points = 5.58 pts/W
9950X, 36W, 143 points = 3.97 pts/W
285K is more efficient in CB 2024 1T by 28%

In review, power consumption in single thread workload is measured in MP3 Encode:
9950X, 58.8s x 36w = 2116.8 J
285K, 66.9s x 26W = 1739.4 J
285K is about 17.2% more efficient than 9950X in MP3 Encode.
So, it depends ...

285K is more efficient than 9950X based on those two calculations above. Valid point there.


Both 285K and 9950X were designed mainly for multithreaded workloads, as their core counts say.
285K was supposed to be beating 9950X thanks to 8 more cores while also being more efficient thanks to node advantage.
That's was I was trying to point out. No valid point there.


720p:
9950X, 104W, 198.6 fps = 1.91 fps/W
285K, 94W, 192.8 fps = 2,05 fps/W
285K is 7.3% more efficient in 720p gaming based on 14 games tested in that review.
Other games = other results. So, it really depends ... But I'll give you another valid point there.

Based on mentioned review, 285K is more efficient than 9950X in: single thread, gaming, idle. In multithread workload not so much, which is a pitty because that's the main area where both CPUs are focusing on (you don't need more than 8 cores for gaming and/or single thread workloads, right?). Still, it would be worth a while to do a re-bench with AMD's updates for Zen 4/5 at place, since in some areas they represent as much as 5-10% performance increase. Another link.

Really interesting lol, really enjoy my 265K tho.