Discussion RDNA4 + CDNA3 Architectures Thread

Page 349 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DisEnchantment

Golden Member
Mar 3, 2017
1,773
6,750
136





With the GFX940 patches in full swing since first week of March, it is looking like MI300 is not far in the distant future!
Usually AMD takes around 3Qs to get the support in LLVM and amdgpu. Lately, since RDNA2 the window they push to add support for new devices is much reduced to prevent leaks.
But looking at the flurry of code in LLVM, it is a lot of commits. Maybe because US Govt is starting to prepare the SW environment for El Capitan (Maybe to avoid slow bring up situation like Frontier for example)

See here for the GFX940 specific commits
Or Phoronix

There is a lot more if you know whom to follow in LLVM review chains (before getting merged to github), but I am not going to link AMD employees.

I am starting to think MI300 will launch around the same time like Hopper probably only a couple of months later!
Although I believe Hopper had problems not having a host CPU capable of doing PCIe 5 in the very near future therefore it might have gotten pushed back a bit until SPR and Genoa arrives later in 2022.
If PVC slips again I believe MI300 could launch before it

This is nuts, MI100/200/300 cadence is impressive.



Previous thread on CDNA2 and RDNA3 here

 
Last edited:

lixlax

Senior member
Nov 6, 2014
202
190
116
Meh, even with this gen, a 96CU-128CU halo part would have done the job IMO. Simply using N3 would have also given them a decent boost as well.
Just to speculate. If the information we know so far is true about RDNA4 then even a not-so-halo 96CU, 384bit, 96MB cache part would've been a decent mindshare boost for AMD. By roughly eyeballing the 4K numbers in reviews it would be anywhere from 0-15% faster than 4090 in raster and probably not very much slower in most RT titles. The die size would've been what- around 450mm2 or a little bit more? They could've even used the same old and cheap 20gbps modules and 24GB would be a nice selling point as well (especially vs RTX5080 and such).
 

Timorous

Golden Member
Oct 27, 2008
1,890
3,638
136
The Nvidia -$50 strategy failed precisely because people acknowledged that Nvidia's exclusive features are worth more than $50. In 2020 and 2022, those were DLSS and (to a lesser degree) RT performance.

Which is why AMD went with a different strategy with the 7000 series. 7900XTX was $200 cheaper than the 4080 for comparable raster performance and a VRAM advantage

The 7800XT was $100 cheaper than the 4070 and was faster in raster and had a VRAM advantage (until NV dropped the price when the 4070S launched). It was the same price as the 4060Ti 16GB despite having a massive performance advantage in raster and being about the same in RT. Still the 4060Ti sold a lot better.

So even when AMD moved to an NV-$100's strategy it still was not enough because the only real reason people want cheap AMD cards is to push NV pricing down so those people can buy a higher tier NV card or get their card for less. The perceived feature disparity is too wide for AMD to bridge.

Even if FSR4 is very comparable to DLSS4 and RT is greatly improved the goalpost will shift to Path Tracing or to the number of available games with FSR4 or to power draw or whatever. There will always be a reason why the AMD card is fundamentally worth less because the reality is people don't want AMD cards, they want cheaper NV cards.

The only way for AMD to break that is to provide a level of performance NV cannot provide at all (like R300 did) and then keep the pressure on gen after gen after gen. They did that from R300 to R580 and you saw their market share during that period, it was very even. The odd stumble won't matter too much if you are reliably providing top tier performance but when you surrender the top end like AMD did with Fiji, then with Vega, then with RDNA1 just going back to competition with RDNA 2 is not enough because of the mind share loss. The surrender also allowed NV to find a way to create feature lock in, first with G-sync and then with DLSS / RT because they could afford to sacrifice some performance to create those features. Look at the way the 2000 series was received and then look at where NVs market share is now. Sure it somewhat hurt them in the short term due to awful PPA and due to the rather meagre performance uplift from 1000 series but now, that bet paid dividends and is allowing NV to control the dGPU market.
 

tsamolotoff

Senior member
May 19, 2019
246
503
136
It was their Nvidia -$50 strategy of RDNA2 and RDNA3 between 2020 and 2024 that killed the mindshare among PC users
RDNA2 price was okay, it was just not available at all due to mining craze. I was trying to buy 6800xt at msrp price on day one, and I couldn't get any (and mining craze hasn't even yet started in full (in countries that are not China, where 5600xts disappeared a few months before the BTC/ETH went sky high), and it wasn't available at non-outrageous price for the next two years or so, despite the fact that n21 wasn't as popular among the e-currency enterpreneurs as Ampere. I guess AMD refused to expand the production after two previous cases where they've bought into the mining craze and got left with warehouses full of unsold Tahiti and Polaris gpus
 
Reactions: Gideon and Tlh97

adroc_thurston

Diamond Member
Jul 2, 2023
5,399
7,570
96
449/499$ 4080s would make all of them move
Either naive or clueless.
(hint: people will wait for NV refreshes/pricecuts).
AMD did gain a truckload of marketshare with Polaris, they went from 20% to 30% in 2 years, and despite the lacking features RDNA1 actually managed to withhold that marketshare a bit.
Loss leader.
Geforce prices are sky high because they're allocating most of their wafers to the AI and top-end chips that get them ridiculous margins
They're sky-high because client Blackwell is a trainwreck that had a late prod kickoff.
They'll easily ramp the supply now.
 

Heartbreaker

Diamond Member
Apr 3, 2006
4,746
6,250
136
Only for them it's need similiar to DLSS Transformer and good looking perfomance mode. Let's hope they tomorrow announce similar to this
In this new video perfomance mode looks better, than old DLSS3 Quality

Only in a couple of categories. In many you need balanced in some you still need Quality.

Overall he considers in one tier better, so you can run balanced now instead of Quality.

Note that Balanced now has performance cost that is higher than it was before.
 
Reactions: Elfear

Mopetar

Diamond Member
Jan 31, 2011
8,314
7,335
136
And they stay in business how?

AMD manages to offset a lot of their costs by partnering with the console manufacturers, so that's a bit part of it. They also have this really successful CPU that you might have heard of.

Though there was a time when the GPU division was carrying the CPU side of the company. Perhaps some day they'll both be firing on all cylinders at the same time.

I don't believe Polaris was financially successful for AMD as they were heavily discounting cards before the crypto boom. 4GB 470s were barely above $100 for a short period. AMD may have sold a lot of cards, but not for as much money as they wanted to.
 
Reactions: Tlh97

gaav87

Senior member
Apr 27, 2024
650
1,267
96
No it was the standard ATi affair of making zero money on the entire lineup.
It is like you are comparing rx5700xt (399$) vs 2070super (499$) slightly worse perf at launch worse features -100$
This time its rx5700xt (399$) vs 2080super (699$) with almost on par features.
Imagine if rx5700xt had 2080s perf back in the day for that price it would be legendary status now.
 
Reactions: Tlh97
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |