Question AMD Rembrandt/Zen 3+ APU Speculation and Discussion

Page 72 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Kaluan

Member
Jan 4, 2022
123
275
96
ComputerBase reviewed different laptops, both Intel and AMD. Please check It out.

What I found very interesting is the Rembrandt's IGP clock speed at different TDP and performance in Cyberpunk.
View attachment 65877

And here is the performance chart.
View attachment 65885

I don't understand why 10W 6800U with IGP clocked <600Mhz performs at 45% of Max(~67W) 6900HS, when that one clocks at ~2300 MHz, which is almost 4x higher.
The second thing I don't understand is why 6900HS 25W scores 10% higher than 6800U 25W, when It has ~15 lower IGP clocks?


edit: Ok, after checking once more, they are both dual channel, but one is using 6400MHz(6800U) and the other 4800MHz DDR5(6900HS). Mystery solved. ;)
Even 45W is not enough for Rembrandt to boost to 2400MHz. It would be interesting to see what's the performance difference between 6600U 25W vs 6800U 25W.
The R9 HS doing <300MHz @ 15W vs R7 U doing >500MHz @ 10W is also a very interesting result. (and being almost 2x faster than the 15W R9 HS in that test)

Shows that U, H, HS and HX parts are defiantly not the same silicon or at least not the same microcode. At least when it comes to <45W test scenarios.

Anyway, I'm a bit confused as to why Rembrandt (especially U class) parts are tested at 25W and not 28W. Is it because of lazy laptop designs or lazy reviewers? You'd think AMD tuned the v/f curves for 28W, not 25W (hence some results may not represent AMD's internal tests/advertising).
 
  • Like
Reactions: Tlh97 and MadRat

moinmoin

Diamond Member
Jun 1, 2017
3,827
5,597
136
The R9 HS doing <300MHz @ 15W vs R7 U doing >500MHz @ 10W is also a very interesting result. (and being almost 2x faster than the 15W R9 HS in that test)

Shows that U, H, HS and HX parts are defiantly not the same silicon or at least not the same microcode. At least when it comes to <45W test scenarios.
Typically U and H series chips should be different binnings, with the former going for better efficiency at lower frequency and the latter for higher leakage allowing for higher frequency. I'd think everything else should be the same, though microcode/bios firmware may contain different behavior tuned specifically to the binning results.
 

MadRat

Lifer
Oct 14, 1999
11,819
154
106
The R9 HS doing <300MHz @ 15W vs R7 U doing >500MHz @ 10W is also a very interesting result. (and being almost 2x faster than the 15W R9 HS in that test)

Shows that U, H, HS and HX parts are defiantly not the same silicon or at least not the same microcode. At least when it comes to <45W test scenarios.

Anyway, I'm a bit confused as to why Rembrandt (especially U class) parts are tested at 25W and not 28W. Is it because of lazy laptop designs or lazy reviewers? You'd think AMD tuned the v/f curves for 28W, not 25W (hence some results may not represent AMD's internal tests/advertising).
I'd look at memory synchronization. They both probably run out of sync with memory speed in some fashion, one more so than the other.
 
  • Like
Reactions: Kaluan

moinmoin

Diamond Member
Jun 1, 2017
3,827
5,597
136
Interesting. Very. What about perf/watt while we're at it ?
That's touched on on the first page:
"Due to the Apple M2 currently lacking any power/temperature sensor support under Linux, this is simply looking at the raw performance of the M2 and Ryzen 7 PRO 6850U with not being able to accurately compare the M2 power efficiency / performance-per-Watt at this time."

Hardware support for M2 under Linux is limited, the GPU isn't supported for example.

But it's a Macbook Air which I'd expect to showcase the same hard throttling as under MacOS.

For the 6850U laptop Michael listed power consumption along test results in an earlier article:
 

guidryp

Platinum Member
Apr 3, 2006
2,392
2,916
136

Apple M2 vs Ryzen 6850u in almost 200 benchmarks. Impressive results for the 6850u, beating the M2 in the Geomean of 190 benchmarks.
Though the M2 only has 4 performance cores, and 8 efficiency cores for 8 threads.

Vs 8 performance cores and 16 threads for the 6850u.

So really I'd would have been surprised if the 6850u didn't win.

I'm kind of surprised M2 did this well, since this is work in progress Asahi Linux running on the M2 that was mostly done with reverse engineered drivers, versus x86 Linux that has been mainstream for decades.
 

DrMrLordX

Lifer
Apr 27, 2000
19,997
8,981
136
Hardware support for M2 under Linux is limited, the GPU isn't supported for example.
It's cool that you can run that much cross-platform software on an M2 machine, but it should be rather obvious that it's suffering for want of properly-compiled binaries.

I'm kind of surprised M2 did this well, since this is work in progress Asahi Linux running on the M2 that was mostly done with reverse engineered drivers, versus x86 Linux that has been mainstream for decades.
From what little I've read, Asahi performance is pretty good except when it comes to the GPU. The problem is more in the software itself and not necessarily the underlying OS.
 
  • Like
Reactions: Lodix and Tlh97

IntelUser2000

Elite Member
Oct 14, 2003
8,383
3,333
136
Looks like AMD has finally figured out a good idling solution. Impressively low idle draw!
Their tests are misleading because that's AC power draw. You'll see many laptops do bad in AC idle but great in battery idle. I doubt battery idle was bad for AMD since Cezanne. Otherwise the battery life would have been bad.

The better way is looking at HWInfo and looking at battery drain data.
 

tamz_msc

Diamond Member
Jan 5, 2017
3,308
3,197
136
We've had dozens of Rembrandt reviews till date post "announcement/launch", but not a single one of them actually tested whether performance of the CPU in your average daily tasks remains consistent when on battery, or whether AMD still ignores Windows performance modes to "win" in battery life comparisons at the expense of performance.
 

Thibsie

Senior member
Apr 25, 2017
432
404
136
That's almost a userbenchmark kind of comment. Did they comment anything on the Intel way of doing boost way past the CPU power rating ?

Well, if you target one, you target both otherwise it stinks...
 

uzzi38

Platinum Member
Oct 16, 2019
2,332
4,840
116
We've had dozens of Rembrandt reviews till date post "announcement/launch", but not a single one of them actually tested whether performance of the CPU in your average daily tasks remains consistent when on battery, or whether AMD still ignores Windows performance modes to "win" in battery life comparisons at the expense of performance.
That is entirely OEM-configurable and even done with some Alder Lake laptops recently as well.

PS: Rembrandt doesn't need it to win on battery life, just FYI.
 
  • Like
Reactions: Kaluan and Tlh97

mikk

Diamond Member
May 15, 2012
3,642
1,511
136
That is entirely OEM-configurable and even done with some Alder Lake laptops recently as well.

PS: Rembrandt doesn't need it to win on battery life, just FYI.

For some reason OEMs make it fast on AC and slow on battery mode on AMD notebooks, performance is more stable on Intel between AC and battery. Performance tests are usually done in AC mode and battery life tests obviously in battery mode. It's a win situation for AMD and OEMs. If this has changed on recent Rembrandt notebooks needs to be tested, unfortunately there is not really a big motivation to find it out.
 

Thibsie

Senior member
Apr 25, 2017
432
404
136
For some reason OEMs make it fast on AC and slow on battery mode on AMD notebooks, performance is more stable on Intel between AC and battery. Performance tests are usually done in AC mode and battery life tests obviously in battery mode. It's a win situation for AMD and OEMs. If this has changed on recent Rembrandt notebooks needs to be tested, unfortunately there is not really a big motivation to find it out.
This is a default behaviour that can be changed by the user.
Do you want to test the CPU or test the OEM ?
The Intel boost behaviour is an easy win for short test. Do anyone yell at Intel 'cos they cheat ?
 

tamz_msc

Diamond Member
Jan 5, 2017
3,308
3,197
136
That is entirely OEM-configurable and even done with some Alder Lake laptops recently as well.

PS: Rembrandt doesn't need it to win on battery life, just FYI.
Intel laptops do not ignore Windows performance settings. AMD laptops do, and that behavior was arguably worse in the past generations.
That's almost a userbenchmark kind of comment. Did they comment anything on the Intel way of doing boost way past the CPU power rating ?
There is no "boosting way past the CPU power rating" with Intel. There's PL1 and PL2. Once you fix the power profile, and provided the thermal design is adequate, only PL1 changes depending on whether you are on battery or AC. PL2 doesn't. What the Windows performance plan changes is the responsiveness of the boost behavior. On AMD while on battery, the latter is slow even if you set it to best performance mode, or at least it used to be.

This is not something you will see in Cinebench runs, because those kind of tests always fall back to PL1 on laptops. You will notice it on browser benchmarks like Speedometer 2.0, Office stuff, light compilation/light compute like MATLAB or Mathematica, and Geekbench.

AMD laptops having better battery life - that's the reason why - the experience on battery is disparate from the experience on AC.
 
  • Like
Reactions: scineram

uzzi38

Platinum Member
Oct 16, 2019
2,332
4,840
116
Intel laptops do not ignore Windows performance settings. AMD laptops do, and that behavior was arguably worse in the past generations.
Windows performance settings have been meaningless for years. The only thing that matters is how the laptops are configured by OEMs, and Intel laptops have very similar knobs as AMD to play around with there.

On AMD while on battery, the latter is slow even if you set it to best performance mode, or at least it used to be.
Like I said before, you get this on some Intel laptops too now - again it depends on how they're configured. I can't remember which laptops I've seen it in, but I can point you towards that Chinese review done by the Lenovo PM we saw very shortly after CES. The Intel laptop there had a worse differential performance on battery than the AMD one.

AMD laptops having better battery life - that's the reason why - the experience on battery is disparate from the experience on AC.
That's a kind of statement that needs to be backed up by actual testing. Why don't you use a tool like AATU that lets you disable the turbo timer, test battery life compared to an identical or near identical Tiger Lake laptop (seeing as Alder Lake is a gen on gen regression).

I can tell you now though, the results won't be what you're expecting.
 

tamz_msc

Diamond Member
Jan 5, 2017
3,308
3,197
136
Windows performance settings have been meaningless for years. The only thing that matters is how the laptops are configured by OEMs, and Intel laptops have very similar knobs as AMD to play around with there.
This is completely false. It takes one look at PC World's testing to disprove your claim.
Like I said before, you get this on some Intel laptops too now - again it depends on how they're configured. I can't remember which laptops I've seen it in, but I can point you towards that Chinese review done by the Lenovo PM we saw very shortly after CES. The Intel laptop there had a worse differential performance on battery than the AMD one.
It's somewhere in this thread, but the Chinese review shows the opposite of what you claim.
That's a kind of statement that needs to be backed up by actual testing. Why don't you use a tool like AATU that lets you disable the turbo timer, test battery life compared to an identical or near identical Tiger Lake laptop (seeing as Alder Lake is a gen on gen regression).

I can tell you now though, the results won't be what you're expecting.
There are few reviews of the 6800U available, but going along with what we have now, Notebook check tested the Zenbook 13 S OLED and the Yoga 9i. The latter with a 4K OLED screen and the i7-1260P. The Zenbook has a 67 Wh battery and the Yoga a 75 Wh. Considering those differences, once normalised the two laptops have comparable battery life while browsing the web.
 
  • Like
Reactions: scineram

Abwx

Diamond Member
Apr 2, 2011
9,755
2,096
136
There are few reviews of the 6800U available, but going along with what we have now, Notebook check tested the Zenbook 13 S OLED and the Yoga 9i. The latter with a 4K OLED screen and the i7-1260P. The Zenbook has a 67 Wh battery and the Yoga a 75 Wh. Considering those differences, once normalised the two laptops have comparable battery life while browsing the web.
Up to 9H for RMB with a 67Wh battery and 7H55 for the 1260P with a 75Wh battery, that s what you call comparable...


 

moinmoin

Diamond Member
Jun 1, 2017
3,827
5,597
136
It's cool that you can run that much cross-platform software on an M2 machine, but it should be rather obvious that it's suffering for want of properly-compiled binaries.
I'm not sure if there's a misunderstanding, a mix up, or you know more than I do. "suffering for want of properly-compiled binaries" is a typical problem with closed source software. Phoronix is using open source software for these benchmarks and lists the specific compiler settings on the first page (the section with white on black text) and , if deviating, as part of the benchmark itself. So either you missed that, or you think the compiler settings are lacking for M2?
 
  • Like
Reactions: Tlh97 and Kaluan

tamz_msc

Diamond Member
Jan 5, 2017
3,308
3,197
136
I suggest you read other responses carefully and examine your own sources more closely in the future. I said Yoga 9i because that also has an OLED screen.


The Yoga 7 that you linked to has an LCD screen and a 71 Wh battery.

Normalise the battery capacity in those comparisons and Alder Lake is slightly behind, but not by much.
 

DrMrLordX

Lifer
Apr 27, 2000
19,997
8,981
136
I'm not sure if there's a misunderstanding, a mix up, or you know more than I do. "suffering for want of properly-compiled binaries" is a typical problem with closed source software. Phoronix is using open source software for these benchmarks and lists the specific compiler settings on the first page (the section with white on black text) and , if deviating, as part of the benchmark itself. So either you missed that, or you think the compiler settings are lacking for M2?
Unless they updated a lot of their applications from the M1 tests, several of them wind up using Rosetta 2.
 

ASK THE COMMUNITY