Speculation: Ryzen 4000 series/Zen 3

Page 64 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

lobz

Platinum Member
Feb 10, 2017
2,057
2,856
136
What about 35W and 45W?
Because the uncore eats up too much power leaving almost nothing to the cpu cores. Even 8 core Epycs uses lot's of power.

I wonder why amd never released some Epyc 5000 series (TR platform based) and Epyc 3000/Opteron 3000 series (am4 platform based). It seams they have a large gap (read market) to fill in servers.
In Renoir the power saved on the I/O die and IF links goes to graphics, so that's always gonna be a stretch either way. Couple of weeks and we'll see :) Here's hoping Vega won't be too much of a drag.
 

Thunder 57

Platinum Member
Aug 19, 2007
2,674
3,796
136
Wow, for more 10W it doubled the clock.

That did seem like an outlier so I tried it again. This time upon restart I let it idle for a bit longer than normal to make sure nothing in the startup process would effect it. It did a bit better. It bounced around a lot between 1.3-1.5GHz. I'd say it averaged around 1.375-1.4GHz during the run. Still a 1.1GHz boost for 10W is crazy. My guess is at that low wattage so much is spent on the IF and memory controller and other necessities that the cores themselves don't have much power to use.
 

uzzi38

Platinum Member
Oct 16, 2019
2,622
5,880
146

amd6502

Senior member
Apr 21, 2017
971
360
136
I'm sorry... what?
I believe _rogame reverse engineered the drivers to some extent and he thinks the CU count is 8 rather than the previous rumors of 12 or 13; I may be wrong though. I just skimmed the info.

Oops (edit), okay going back this thread (post #1551) it seems rogame is uzzi's "guy" not yours; sorry for my confusion.
 
Last edited:
  • Like
Reactions: lobz

lobz

Platinum Member
Feb 10, 2017
2,057
2,856
136
I believe _rogame reverse engineered the drivers to some extent and he thinks the CU count is 8 rather than the previous rumors of 12 or 13; I may be wrong though. I just skimmed the info.

Oops (edit), okay going back this thread (post #1551) it seems rogame is uzzi's "guy" not yours; sorry for my confusion.
Man, when the big time wall street guy aka liahos mistook me for @uzzi38 , I thought damn if it's that easy to steal someone's identity. :D
 

uzzi38

Platinum Member
Oct 16, 2019
2,622
5,880
146
I believe _rogame reverse engineered the drivers to some extent and he thinks the CU count is 8 rather than the previous rumors of 12 or 13; I may be wrong though. I just skimmed the info.

Oops (edit), okay going back this thread (post #1551) it seems rogame is uzzi's "guy" not yours; sorry for my confusion.
Yep, you're correct. That's almost exactly what he did. What he found was pretty clear.

Arcturus is 8 Shader Engines with 16CUs each (128CUs total)
Navi12 is 2 Shader Engines with 10WGPs each (40CUs total)
Renoir is 1 Shader Engine with 8CUs in it.
 
  • Like
Reactions: amd6502

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
It probably means a variant of Renoir has 8CUs, rather than that really being the maximum. That's why other numbers float around.
 

soresu

Platinum Member
Dec 19, 2014
2,657
1,858
136
Navi12 is 2 Shader Engines with 10WGPs each (40CUs total)
Lends credence to the Apple only Navi10 HBM variant rumour?

Unless Navi10 also had significant errata now fixed, or pieces missing now added to Navi12 that makes it viable giving it a new name beyond HBM2 support alone.
 

uzzi38

Platinum Member
Oct 16, 2019
2,622
5,880
146
It probably means a variant of Renoir has 8CUs, rather than that really being the maximum. That's why other numbers float around.
No, he looked at firmware. What he found is in regards to the full die.

However, we both believe it may not be accurate as again, rumour suggests otherwise. And we asked several people up the rumour foodchain as well, who also said it didn't line up with what they'd heard. Feel free to disregard it.
 

uzzi38

Platinum Member
Oct 16, 2019
2,622
5,880
146
Lends credence to the Apple only Navi10 HBM variant rumour?

Unless Navi10 also had significant errata now fixed, or pieces missing now added to Navi12 that makes it viable giving it a new name beyond HBM2 support alone.
HBM2 is basically confirmed. Navi 12 is almost certainly mobile focused too.


But Navi 10 also did have errata that is being adressed with Navi 12.


(SDMA stands for System Direct Memory Access).
 
  • Like
Reactions: amd6502 and soresu

uzzi38

Platinum Member
Oct 16, 2019
2,622
5,880
146
I wonder if this will lead to extra efficiency or performance beyond HBM2.

Well, I can't say I know what SDMA is actually for, but if I'm gonna take a guess from the name, whatever this errata is, I get the feeling it's directly related to the PCIe Gen 3 vs PCIe Gen 4 weirdness surrounding the 5500XT.
 

H T C

Senior member
Nov 7, 2018
549
395
136
Well, I can't say I know what SDMA is actually for, but if I'm gonna take a guess from the name, whatever this errata is, I get the feeling it's directly related to the PCIe Gen 3 vs PCIe Gen 4 weirdness surrounding the 5500XT.

There is no weirdness.

What happens is that in ALL PCIe 3.0 boards, the 5500XT works @ x8 PCIe 3.0 while with X570 boards it works @ x8 PCIe 4.0. The difference stems from the fact that boards with PCIe 3.0 have HALF the bandwidth of X570 boards, that have PCIe 4.0.

Effectively, when you place ANY PCIe 4.0 card in a non-PCIE 4.0 board, you're bottlenecking the card's true performance. It the case of the 5700(XT) card(s) you don't notice it much because of the VRAM present in the card but you may notice in SOME games if you use a 5500(XT) card.