Question Speculation: RDNA2 + CDNA Architectures thread

Page 212 - 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
2,698
6,393
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
 

Mopetar

Diamond Member
Jan 31, 2011
8,000
6,433
136
Based on the MSRP price, the 6700 XT needs 82.75% of the performance of the 6800 to be worth purchasing in terms of value/dollar. In terms of processing power, the faster clock speeds put it in line with or above where it needs to be to deliver on that and unless it's bottlenecked by ROPs or memory bandwidth it should be able to deliver at 1080p especially as well as 1440p.
 

Timorous

Golden Member
Oct 27, 2008
1,726
3,141
136
That depends. Certainly Navi10 level of performance for N23 would be best, but new features and much lower TBP could be worth It especially in mobile segment where AM|D is not faring very well. Smaller size also helps lowering production costs and allowing to produce more.

It just seems that 5700XT should be the lower bound for a desktop RDNA2 GPU at 230mm². Realistically if the desktop version clocks as high as the 6700XT you would think it should slightly outperform the 5700XT at 1080p.
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,414
2,906
136
It just seems that 5700XT should be the lower bound for a desktop RDNA2 GPU at 230mm². Realistically if the desktop version clocks as high as the 6700XT you would think it should slightly outperform the 5700XT at 1080p.
It depends on what AMD aims for with N23.
If they want performance, then they will clock It as high as RX 6700XT and at 2.5Ghz It would have more brute force(TFlops) than RX 5700XT.
If they prefer efficiency and low TBP, then they will set conservative clocks comparable to RX 6800.
 

Timorous

Golden Member
Oct 27, 2008
1,726
3,141
136
It depends on what AMD aims for with N23.
If they want performance, then they will clock It as high as RX 6700XT and at 2.5Ghz It would have more brute force(TFlops) than RX 5700XT.
If they prefer efficiency and low TBP, then they will set conservative clocks comparable to RX 6800.

They can do both with the 6600XT being high TBP and the 6600 or even a 6500XT using the same chip being significantly lower clocks for the TBP savings.
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,414
2,906
136
N23 is for Full HD. N24 will be weaker and more suitable for mobile, like HD Ready, potentially replaced by future APUs.
With N23 at RX5700XT level of performance you can play Control at 1440p ~40FPS at highest settings.
N24 should have at least 20CU and with higher clocks than RX5500XT It will perform better and Navi14 is capable of playing Control in Full HD ~35FPS at highest settings.
Control is the most demanding game in TPU reviews so far.
So I don't understand why It should be suitable only for HD Ready, I think It's enough even for FullHD.
 

Timorous

Golden Member
Oct 27, 2008
1,726
3,141
136
Can you show me some example of that happening from previous generations?
Not to mention I though we were talking about standard retail models and not something aimed at OEMs.

You brought up low power. Personally I think AMD will clock it as high as they can because there is no reason not to.
 

Glo.

Diamond Member
Apr 25, 2015
5,759
4,666
136
That low power(TBP: 135W) comes from glo.
Don't know If It's still valid.
Of course it is.
N23 is for Full HD. N24 will be weaker and more suitable for mobile, like HD Ready, potentially replaced by future APUs.
With desktop class clocks, like 2.35 GHz, between GTX 1660 Ti and RTX 2060 performance level is enough for Mobile applications, and will be easily replaced by future APUs?

 

Timorous

Golden Member
Oct 27, 2008
1,726
3,141
136
That low power(TBP: 135W) comes from glo.
Don't know If It's still valid.

Seems hitable when the entite PS5 uses around 205W in games. Fewer CUs, half the ram, half the memory bus width and no CPU, SSD etc to power.

Even a 5% clock reduction will reduce power significantly and the 2.3Ghz I estimate is needed for 32CUs to match the 5700XT is a 5% reduction from the 6700XT game clock.
 

Glo.

Diamond Member
Apr 25, 2015
5,759
4,666
136
Seems hitable when the entite PS5 uses around 205W in games. Fewer CUs, half the ram, half the memory bus width and no CPU, SSD etc to power.

Even a 5% clock reduction will reduce power significantly and the 2.3Ghz I estimate is needed for 32CUs to match the 5700XT is a 5% reduction from the 6700XT game clock.
N23 is slightly faster than RX 5700 XT.
 

Mopetar

Diamond Member
Jan 31, 2011
8,000
6,433
136
Having 80% of the CUs only requires clocks to be 25% faster to get back to parity, and we know that's certainly possible.

I wonder how much performance falls off past 1080p though, especially if the amount of cache is smaller as rumored. The 128-bit bus is already half the size as the 5700 XT which makes me think that it may not be quite as good at 1440p, but that's not an issue if it's priced as the entry level card. However, with the 6700 XT at $480, I don't see even the entry-level card having an entry-level price.
 

Timorous

Golden Member
Oct 27, 2008
1,726
3,141
136
Having 80% of the CUs only requires clocks to be 25% faster to get back to parity, and we know that's certainly possible.

I wonder how much performance falls off past 1080p though, especially if the amount of cache is smaller as rumored. The 128-bit bus is already half the size as the 5700 XT which makes me think that it may not be quite as good at 1440p, but that's not an issue if it's priced as the entry level card. However, with the 6700 XT at $480, I don't see even the entry-level card having an entry-level price.

I expect it will be the same price as the 5700 was at launch. Slightly smaller die, slightly better perf than 5700X if Glo is correct so you get a mild perf/$ uplift and AMD has similar or slightly better margins on it.
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,414
2,906
136
Seems hitable when the entite PS5 uses around 205W in games. Fewer CUs, half the ram, half the memory bus width and no CPU, SSD etc to power.

Even a 5% clock reduction will reduce power significantly and the 2.3Ghz I estimate is needed for 32CUs to match the 5700XT is a 5% reduction from the 6700XT game clock.
Of course It's doable, lowering clocks and voltage does magic. The question is If AMD didn't change their mind and instead of efficiency goes for very high clocks, which would noticeably increase TBP.
As Glo. pointed out, It's still aiming for efficiency.
 

Timorous

Golden Member
Oct 27, 2008
1,726
3,141
136
Of course It's doable, lowering clocks and voltage does magic. The question is If AMD didn't change their mind and instead of efficiency goes for very high clocks, which would noticeably increase TBP.
As Glo. pointed out, It's still aiming for efficiency.

RDNA2 seems okay with both. Not expecting 2.424 Ghz but 2.3Ghz for a GPU is very high and prior to RDNA2 people would have laughed at you.
 

insertcarehere

Senior member
Jan 17, 2013
639
607
136
Seems hitable when the entite PS5 uses around 205W in games. Fewer CUs, half the ram, half the memory bus width and no CPU, SSD etc to power.

Even a 5% clock reduction will reduce power significantly and the 2.3Ghz I estimate is needed for 32CUs to match the 5700XT is a 5% reduction from the 6700XT game clock.
The PS5's GPU isn't full RDNA2, some features are missing (HW VRS) while other units are divergent from Navi 2X (No Infinity Cache, different ROP layout (similar to RDNA 1), so making direct comparisons is not nearly as simple.
 

Mopetar

Diamond Member
Jan 31, 2011
8,000
6,433
136
How much would those change the expected power use though? Infinity cache is supposed to be more power efficient than a wider bus, so we should expect Navi 23 to have better power characteristics there.

Maybe some of the customizations that Sony made help balance it out because they're more efficient, but the clock speeds and voltage are going to play a bigger factor than all of the small differences between whatever the PS5 has and what RDNA2 had.
 

Timorous

Golden Member
Oct 27, 2008
1,726
3,141
136
The PS5's GPU isn't full RDNA2, some features are missing (HW VRS) while other units are divergent from Navi 2X (No Infinity Cache, different ROP layout (similar to RDNA 1), so making direct comparisons is not nearly as simple.

It is an estimate using a measured and broadly similar product to make a ballpark guess as to what is possible.

The PS5 and Series X clocks / PSU size indicated that RDNA2 would clock well, guess what it does. In a similar fashion the combination of PS5 measured power consumption and released 6700XT specs indicates that a 32CU, 128bit, 32MB cache RDNA2 part should have good power consumption figures (sub 150TBP) if it is backed off slightly from the 2.424 Ghz clock of the 6700XT.
 

GodisanAtheist

Diamond Member
Nov 16, 2006
7,039
7,461
136
I honestly miss the point of Navi24. Entry level seems already covered well by Navi23.

-I suspect it's something that was designed for Apple or to fit some design parameters for a large OEM and in classic AMD form they won't let any custom design work go to waste and we'll see the chip shoehorned into a discreet design.

From the looks of it, AMD would have been able to cover the entire range with an 80/40/20 CU layout + respective 72/32/12CU cut down designs.
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,414
2,906
136
I honestly miss the point of Navi24. Entry level seems already covered well by Navi23.
From when is performance similar to RX 5700XT entry level?
I think the price is the biggest problem.
RX 6700XT's MSRP is $479. So how likely is that any variant of N23 will cost less than $299?
Most gamers are not willing to spend $300 or more on a graphic card.
 
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/    |