Question Speculation: RDNA2 + CDNA Architectures thread

Page 88 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

uzzi38

Platinum Member
Oct 16, 2019
2,705
6,427
146
All die sizes are within 5mm^2. The poster here has been right on some things in the past afaik, and to his credit was the first to saying 505mm^2 for Navi21, which other people have backed up. Even still though, take the following with a pich of salt.

Navi21 - 505mm^2

Navi22 - 340mm^2

Navi23 - 240mm^2

Source is the following post: https://www.ptt.cc/bbs/PC_Shopping/M.1588075782.A.C1E.html
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
PP tables are the same everywhere for the same HW.
PP tables are soft limits. You can overwrite them, like how Igor is doing with his tool.
Hard limits are in the atombios.
Even in Apple hardware? Those specs are specifically for GPUs that will go to Apple computers. Somewhat like LInux GPU microcode.
 

DisEnchantment

Golden Member
Mar 3, 2017
1,747
6,598
136
Even in Apple hardware? Those specs are specifically for GPUs that will go to Apple computers. Somewhat like LInux GPU microcode.
Those are not specific to Apple, firmware blob is same everywhere for the same HW/SKU.
Apple gets it own SKU, which has its own firmware(and PP table).
Afterwards OEM can change whatever (PP table part) they want as long as it is with the limits defined inside atombios.
The firmware has microcode as well which instructs how the different ip blocks will function. Like CPU microcode.

Firmware is provided by AMD.
AMD provides an interface via driver to overwrite the PP table.
 
Last edited:

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
I saw the power tables. Might not have anything to do with apple as the OP says.
So it’s an unknown right now it could be potentially another hundred watts added on.
Looking at Navi 10 and Navi 14 power tables, not really that is the case.

It appears that most likely picture is that 238W and 170 is the actual power target for those GPUs.
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
possible, but as the OP says N21 is really still unknown. Exciting though so far.
The thing is.

In the clock tables we see that N21 has 2050 MHz clock target, while maximum is 2200 MHz. 2050 is the "gameclock" clock, and 2200 MHz is the boost clock?

If so, why N22 does not have "gameclock" and has exactly the same clock table as maximum Boost clock?

Also, Even 238W at 2050 MHz for whole board in 80 CU GPU is pretty darn impressive efficiency achievement.

It basically means that AMD achieved at least 2x the efficiency, on the same node, if those clock and power tables are going to have any relation to the real world performance.

2x performance/watt. On the same node.

O_O
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,523
3,037
136
Another weird thing is that N22 has 40CU, 192bit 16GHz GDDR6 and N23 has 32CU but only 128bit 14-16Ghz gddr6. N22 should be 48CU, then It would be 50% difference in CU and bandwidth. That would make much more sense and there wouldn't be so much difference between N21 and N22.
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
Another weird thing is that N22 has 40CU, 192bit 16GHz GDDR6 and N23 has 32CU but only 128bit 14-16Ghz gddr6. N22 should be 48CU, then It would be 50% difference in CU and bandwidth. That would make much more sense and there wouldn't be so much difference between N21 and N22.
What if:
N21 - 80 CUs
N22 - 40 CUs
N23 - 32 CUs
N24 - 64 CUs

Doesn't this make sense?
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,523
3,037
136
What if:
N21 - 80 CUs
N22 - 40 CUs
N23 - 32 CUs
N24 - 64 CUs

Doesn't this make sense?
64CU can be cut down from N21, you don't need a new chip for It.
N22 and N23 is too close.
I think
N21 - 80CU 384bit
N22 - 48CU 192bit
N23 - 32CU 128bit
Would make more sense.
BTW at reddit Navi 10a and Navi 10b has only 300-1000Mhz or 1400 in freq_table_gfx[1] (MHz) while navi14 has 300-1900mhz for some reason.
 

moinmoin

Diamond Member
Jun 1, 2017
5,064
8,032
136
Yep. Its worth adding those are GPUs for Apple, and those usually have lower power and lower clock speeds, compared to desktop GPUs.
There is nothing Apple-specific about the specs. The most surprising part is that Apple keeps including (or AMD keeps giving them) the hardware data for the APUs. With the upcoming transition to Apple Silicon the chance of AMD APUs actually being used in Apple hardware is essentially nil.
OP here. These pptables are test configurations, they are not used by production hardware (you can likely boot with some parameters to use them though). And yes they are the same in macOS and in Linux.
Somebody asked and you answered this already on reddit, but I'd like to second the request reading this data from earlier macOS versions to compare how much changed, if at all.

But in any case this is a great find, thanks for sharing (especially also for documenting how to reproduce)!
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,523
3,037
136
Socket power limit for Navi 10a and b is only 180W and for Navi 14 110W, I think It's just TDP and not TBP.
So 170 or 240W won't be for the complete board. With that It looks like 50% better performance/W is most likely true and 2x better performance/W was BS as I said many times.
 
Last edited:

uzzi38

Platinum Member
Oct 16, 2019
2,705
6,427
146
There is nothing Apple-specific about the specs. The most surprising part is that Apple keeps including (or AMD keeps giving them) the hardware data for the APUs. With the upcoming transition to Apple Silicon the chance of AMD APUs actually being used in Apple hardware is essentially nil.

Somebody asked and you answered this already on reddit, but I'd like to second the request reading this data from earlier macOS versions to compare how much changed, if at all.

But in any case this is a great find, thanks for sharing (especially also for documenting how to reproduce)!
Pretty sure AMD just gives them the driver code as one big blob and they just integrate it in.

Nothing more to it.

More importantly, did anyone notice how the Navi10 clocks are extremely low?
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,706
1,233
136
N21 = 80 CUs
N22 = 40 CUs
N23 = 24 CUs
N24 = 10 CUs

ETYloJ3WsAEHpqM.png

arden.jpg
Thumbnail click it:
4x14 in Arden, sorry about the weird counting I did.

vangogh.jpg
Thumbnail click it:
3x8 in Mero which is also Van Gogh. Normal counting here, sorry.

=> 00890F00h \ Zen2 \ VanGogh \ VN-A0 \ 7nm
=> 00890F80h \ Zen2 \ Mero \ MR-A0 \ 7nm
Same model range = same die.
Van Gogh being Family 17h 90h-97h and Mero being Family 17h 98h-9Fh.

It puts GFX1032 specs in question as well since only Navi21 SCBU is different than Navi21 actual.

Arden is Navi21 SCBU => 80 CUs to 56 CUs
Ariel is Navi22 SCBU => 40 CUs to 40 CUs
Mero is GFX1032/GFX1033/GFX1040 which is Navi23 SCBU => 24 CUs to 24 CUs.

We have to confirm from two sources and more these days. Or, we are going to get dumped with 7.5W-18W Van Gogh rumors... :laughing:

The full stack part is this:
N22 => P20
N23 => P22
N24 => P23
 
Last edited:

uzzi38

Platinum Member
Oct 16, 2019
2,705
6,427
146
Socket power limit for Navi 10a and b is only 180W and for Navi 14 110W, I think It's just TDP and not TBP.
So 170 or 240W won't be for the complete board. With that It looks like 50% better performance/W is most likely true and 2x better performance/W was BS as I said many times.
This. The power figures there are very likely GPU only.

This is a huge guess, but I'm expecting 300W TBP at that higher power target and ~275W at the lower one.
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
So 170 or 240W won't be for the complete board. With that It looks like 50% better performance/W is most likely true and 2x better performance/W was BS as I said many times.
Thats not how I remember this from the time that I was more interested in Apple software, and kexts.

I'll check on that, still not saying youre wrong, tho.
 

jpiniero

Lifer
Oct 1, 2010
15,223
5,768
136
There is nothing Apple-specific about the specs. The most surprising part is that Apple keeps including (or AMD keeps giving them) the hardware data for the APUs. With the upcoming transition to Apple Silicon the chance of AMD APUs actually being used in Apple hardware is essentially nil.

It's entirely possible that there will be an x86 refresh of the 21" iMac, just like the 27" did in August. While the odds of that being an AMD APU is pretty low, it is possible.
 

Glo.

Diamond Member
Apr 25, 2015
5,803
4,777
136
Oh, you may be right actually.

Take a look at this:


And here is full Radeon Pro 5700 XT specs. So those might be power tables for Apple GPUs, after all, and Im remembering this correctly ;).
 

CastleBravo

Member
Dec 6, 2019
120
271
136

And here is full Radeon Pro 5700 XT specs. So those might be power tables for Apple GPUs, after all, and Im remembering this correctly ;).

I wonder if Apple is choosing to run Navi21 a bit slower than what we might see in the PC market...
 
  • Wow
Reactions: lightmanek

TESKATLIPOKA

Platinum Member
May 1, 2020
2,523
3,037
136
Thats not how I remember this from the time that I was more interested in Apple software, and kexts.

I'll check on that, still not saying youre wrong, tho.
I am just saying based on what was shown in the tables, I don't know If It's correct or not.
Just for comparison purposes.
20190612-graphics-card-tdp-and-tgp.jpg

Navi22 could be TDP: 170W and TBP: 200-215W. Don't know why 5700 has 30W and XT has 45W increase over TDP for TBP, when the memory is the same, that's why I used both for Navi22.
 
Last edited: