Question Speculation: RDNA2 + CDNA Architectures thread

Page 127 - 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
2,565
5,570
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
 

Panino Manino

Senior member
Jan 28, 2017
813
1,010
136
So, no AIB cards at launch?!

Isn't because despite having the chip on their hands they can't see and understand what's inside?
Also, is AMD the one that makes the driver and bios which reveals the features, right? And AMD is trying to give them these codes as late as possible, so there's no way to know for certain outside of AMD.
 

PhoBoChai

Member
Oct 10, 2017
119
389
106
Isn't because despite having the chip on their hands they can't see and understand what's inside?
Also, is AMD the one that makes the driver and bios which reveals the features, right? And AMD is trying to give them these codes as late as possible, so there's no way to know for certain outside of AMD.

Exactly what NV did with Ampere, but still so much leaked out.
 

zinfamous

No Lifer
Jul 12, 2006
110,509
29,090
146
Well, they already said there is an "XTX," and I just love that this nonsense is being resurrected. Hopefully, it means a true XT > XTX > XTXTX > XTXXTTTXXX sort of thing and that there is real performance (like actually used to happen) attached to the nonsense (and so it is defensible), but I doubt it is more than a 3090 competitor, whatever that means (price, performance, efficiency, etc.)

and whenever it arrives, with whatever Nvidia has to throw it at that unannounced time.
 
  • Like
Reactions: kurosaki

Saylick

Diamond Member
Sep 10, 2012
3,084
6,184
136
Well, they already said there is an "XTX," and I just love that this nonsense is being resurrected. Hopefully, it means a true XT > XTX > XTXTX > XTXXTTTXXX sort of thing and that there is real performance (like actually used to happen) attached to the nonsense (and so it is defensible), but I doubt it is more than a 3090 competitor, whatever that means (price, performance, efficiency, etc.)

and whenever it arrives, with whatever Nvidia has to throw it at that unannounced time.
I need XFX to come out with a RX6900XTX XTRA THICCC Pro III triple slot cooler version.
 

gorobei

Diamond Member
Jan 7, 2007
3,654
980
136
Maybe because Nvidia have more works that could leak information?
nvidia has pissed off more of its partners that could leak.

  • direct sales of founders edition cannibalizing aib sales to pad nv profits.
  • board partners forced to buy pascal mining craze overstock when eth crypto bubble burst in order not to get "de-prioritized" in allotments of turing.
  • nv partner program taking over control of gaming branding names (rog, aorus, etc)
  • being forced to submit bom of cooler costs to nv for authorization. just to make sure the cooler isnt better than the FE.
 

ksosx86

Member
Sep 27, 2012
105
44
101
NVIDIA has pissed off more than partners. They soured their relationship with TSMC which is part of why they went to Samsung (just word on the street I've been hearing where I work). Aside from anything they may have said in the press or regarding technical reasons for the architectural designs of Ampere and Samsung UEV 8nm etc. Anyhoo - not that it's anything "confidential" or not just talk - regarding AMD honestly I think they're at 3080 performance and have been using up until now to attempt to tweak the Navi24 to compete with 3090 if possible far as clocks and combinations of memory, bus and such and so forth. Although I'm not sure when one may see the early stages of those results could be Q1 of next year?

I think if I had my product ready for the Holidays had scaled and began production up until now to match projected demand I would only be able to modify anything and I mean anything through severely and I mean severely quality assured UEFI updates if anything is changed this late in the game. At best probably 3080. Beyond that - HBM2E + another Navi24 for Q1 2021 if it exists as it's speculation. Might get 3090 performance out of that (honestly likely more than). I'd be surprised if AMD pulls a Ryzen with RDNA2 and comes out with RDNA2 scaled all the way up to Navi24 matches (well and then some) the 3090 at launch with Raytracing and just kills the Ampere launch. That would take incredible foresight. They have the money now. (At least the credit/financing)
 

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
Navi24 is a small chip. It will likely be what is used to kill off polaris.

I don’t buy the rumor that there is an >80CU chip unless RDNA2 scales better with CU count then it does with frequency.
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Navi24 is a small chip. It will likely be what is used to kill off polaris.

I don’t buy the rumor that there is an >80CU chip unless RDNA2 scales better with CU count then it does with frequency.

Yes. Navi 24 is most likely 16-20 CU and will finally retire Polaris after almost 5 years of service. Polaris is probably the longest lived GPU from either GPU vendor in history.
 

Glo.

Diamond Member
Apr 25, 2015
5,657
4,409
136
Is this the point where I mention how the Apple firmware stuff was pre-final silicon?

I'll be honest, I didn't think clocks would go this high though.
Or simply in Apple firmware we had actual data for Apple only Radeon Pro SKUs.

If 72 CU from N21 XT clocks around 2.4 GHz GAME CLOCK at 255 TDP, I see zero problem for full, 80 CU to clock 2200 MHz at 238W of power.

Wasn't the 2.7 GHz boost clock leaked already?
Yes it was.

And it still appears to be completely ludicrous thought.
 
  • Like
Reactions: Mopetar and Tlh97

eek2121

Platinum Member
Aug 2, 2005
2,904
3,906
136
VideoCardz still thinks this card will trail the 3080. I call redacted. Assuming that there are no other scaling issues, an 80CU card with a 1.825 Game Clock would equal the 3080 in performance. The fact this 72CU card can hit 2.4 GHz means it will walk all over the 3080.





No profanity allowed in the tech forums


esquared
Anandtech Forum Director
 
Last edited by a moderator:

Karnak

Senior member
Jan 5, 2017
399
767
136
VideoCardz still thinks this card will trail the 3080. I call bullshit. Assuming that there are no other scaling issues, an 80CU card with a 1.825 Game Clock would equal the 3080 in performance. The fact this 72CU card can hit 2.4 GHz means it will walk all over the 3080.
Yeah I don't know what whycry's problem is. It's obvious that IF these specs are true then it's more than just his "beats the 3070".
 

Glo.

Diamond Member
Apr 25, 2015
5,657
4,409
136
72 CUs, 255W TDP, 2.4 GHz Game Clock is too good to be true... That's literally like 2x perf/W. I'm not going to keep my hopes up for that tbh.
That is more than 2X performance, already.

And excluding any performance/clock uplift this architecture could bring...

Any IPC performance per FLOPs uplift over RDNA2, and we only increase the performance uplift, ON THE SAME FREAKING NODE!
VideoCardz still thinks this card will trail the 3080. I call bullshit. Assuming that there are no other scaling issues, an 80CU card with a 1.825 Game Clock would equal the 3080 in performance. The fact this 72CU card can hit 2.4 GHz means it will walk all over the 3080.
Who cares what VCZ thinks?