• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Question Speculation: RDNA2 + CDNA Architectures thread

Page 198 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Glo.

Diamond Member
Apr 25, 2015
4,657
3,282
136
Getting back to SAM tech, it appears that Asus's 400 series Intel MoBo's already have RBAR support, for RDNA2 GPUs.


So this will be perfect option for RKL-S CPUs.

For which I am very happy about :)
 

Glo.

Diamond Member
Apr 25, 2015
4,657
3,282
136
I don't doubt that they'll have some version of Navi 23 that's also had the clock speeds pushed as far as possible, but I don't think that necessarily matters for the point of this discussion.

Just look at a site like Silicon Lottery to see what I'm getting at since they have posted statistics about how as you push clock speeds higher, fewer and fewer chips can reach those levels. Let's just use their results for the 10900K (comet lake) to illustrate this:

Clock SpeedPercent of CPUs Capable
4.8 GHz100%
4.9 GHz99%
5.0 GHz68%
5.1 GHz21%
5.2 GHz1%

Some dies are going to have defects in the shaders, etc. that mean they get binned to a 36CU part regardless of performance, but we know that TSMC has really good 7nm yields and that most (80% plus) of the dies AMD will be getting back will be good dies. Depending on the economics how many cards you can expect to sell at a particular price you may still bin some otherwise functional dies just because you've saturated the demand for the highest performance and priced part.

So let's just imagine a hypothetical scenario with Navi 22. I'll just use the same spread as with 10900K since I have nothing else to go off of or a better starting point. It's probably not accurate, but this is just for illustration.

Clock SpeedPercent of GPUs Capable
1.8 GHz100%
2.1 GHz99%
2.4 GHz68%
2.7 GHz21%
2.85 GHz1%

Suppose you wanted to target 2.7 GHz for your 6700XT. Only 21% of your full dies make that cut, so if you want the 6700 to be a 36 CU part you're actually binning 80% of your dies even though they could otherwise still hit reasonable clock speeds that where your previous intention in the first place. If you had stayed at something like 2.2 GHz, you might only need to bin 10% of those full die chips.

But you can have both. The top 20% of chips, go into a 6700XT at 2.7 GHz. The next 70% go into a 6700 at 2.2 GHz, and the bottom 10% get the worst performing CUs disabled to allow them to clock higher as a 6600-level part. They can obviously choose other points along that curve to get the kinds of numbers that they need based on sales expectations, etc.

The point about bringing up Navi 23 was that since they'll have another die that will top out at 32 CU, they probably don't have as much incentive to bin 80% of their full Navi 22 dies because they need cards with 32 or 36 CU to target the $300 - $375 part of the market.

This does make the pricing a little more interesting. If AMD does have a Navi 22 that equals the 3070 is 1080p and 1440p due to high clock speeds what do you think they charge for it? I can see an argument for $500, but something like $470 seems more likely. They also need to leave room for whatever they'll have at $400 to square off against the 3060 Ti.
When I talked about 40 CU die being 10% faster than RTX 2080 Super it was with higher Max boost clock than Navi 21, which at that time it was 2.1 GHz(!). If that 10% above RTX 2080 Super came from 2.5 GHz expect that AMD could've pushed this to 2.7 GHz, even, and 225-240W total board power.

Cut down N22 will be in the range of 175-180W of power.
 

Mopetar

Diamond Member
Jan 31, 2011
5,254
1,866
136
So 2 months before review are out. Can't say I am happy about that.
CES is in early January and they could announce the products for Navi 22 at that. It's such an easy talking point for them at a big event. A ~$400 card that can probably get close to a 2080 Ti in the resolutions it targets is a hell of a lot more exciting to talk about than the Radeon VII.
 

TESKATLIPOKA

Senior member
May 1, 2020
299
315
96
CES is in early January and they could announce the products for Navi 22 at that. It's such an easy talking point for them at a big event. A ~$400 card that can probably get close to a 2080 Ti in the resolutions it targets is a hell of a lot more exciting to talk about than the Radeon VII.
RTX3060Ti reviews were released today and I consider that card to be the most likely competitor for N22. I am more excited about N23, but I don't think price will go under $200.
 

TESKATLIPOKA

Senior member
May 1, 2020
299
315
96
N24 -> 24CU + 64bit GDRR6 and most likely 32MB IC. That should provide a bit more than 50% hitrate, I hope there won't be a bottleneck considering It has a higher number of CU per IC and bus width vs N21(Big Navi). If they manage high enough clocks to make It at least 20% faster than RX 5500XT within 75W TBP, put 8GB Vram and It will cost at most $179, It could be a pretty good lowend GPU.
 
Last edited:
  • Like
Reactions: Tlh97

Mopetar

Diamond Member
Jan 31, 2011
5,254
1,866
136
I wasn't aware there was an N24 die as well, but I suppose it makes sense to have a small die like that. However it does make the inclusion of Navi23 really weird with it being sandwiched halfway in-between a 24 and 40 CU part. Is Navi24 not meant to have Infinity Cache or is there some other reason for AMD to create a lineup like that?
 

moinmoin

Platinum Member
Jun 1, 2017
2,234
2,659
106
Is Navi24 not meant to have Infinity Cache or is there some other reason for AMD to create a lineup like that?
I'd expect one out of N23 and N24 to be the GPU chip optimized for low power mobile usage, the other the low cost low budget chip.
 

Mopetar

Diamond Member
Jan 31, 2011
5,254
1,866
136
I'd expect one out of N23 and N24 to be the GPU chip optimized for low power mobile usage, the other the low cost low budget chip.
Probably the Navi23 die if that's the case. Navi24 looks like a Navi14 replacement since it's pretty similar to it in the same way Navi22 is a Navi10 replacement.
 

Glo.

Diamond Member
Apr 25, 2015
4,657
3,282
136
I'd expect one out of N23 and N24 to be the GPU chip optimized for low power mobile usage, the other the low cost low budget chip.
Both are mobile first designs.

There is a good reason why those GPUs have Infinity Cache, and very small GDDR6 buses.

96 and 128 bit GDDR6 are 15 and 20W of power, respectively. Also require only 3 and 4 memory chips which lowers manufacturing costs, and simplifies the board designs.

I would personally expect that N24 will fit in sub-50W thermal envelopes, while N23 - sub 100W thermal envelopes.
 

soresu

Golden Member
Dec 19, 2014
1,571
784
136
6900XT will be sold by AIBs AND will have a max OC of 3Ghz
If the 3 Ghz run of 6900XT on Firestrike is accurate then that clearly isn't a rumour, unless it's a custom board BIOS only trick.

As for aftermarket/custom 6900XT's happening, I thought that was laid to rest weeks ago, not long after RX 6000 launch?
 

soresu

Golden Member
Dec 19, 2014
1,571
784
136
I would personally expect that N24 will fit in sub-50W thermal envelopes, while N23 - sub 100W thermal envelopes.
I've been wondering if the expected GPU for Raphael is RDNA2 or RDNA3.

If the former then it seems highly likely that one of those is intended for Raphael.
 

Ajay

Diamond Member
Jan 8, 2001
7,925
2,973
136
If the 3 Ghz run of 6900XT on Firestrike is accurate then that clearly isn't a rumour, unless it's a custom board BIOS only trick.

As for aftermarket/custom 6900XT's happening, I thought that was laid to rest weeks ago, not long after RX 6000 launch?
Got to be at least liquid cooled.
 

TESKATLIPOKA

Senior member
May 1, 2020
299
315
96
All I can tell you is that you can expect at least GTX 1660 Ti +15% performance, with this small, cheap GPU.
That would mean ~40% over RX 5500XT, which is pretty close to RTX 2060 6GB, but the amount of Vram is just not good and It copies Nvidia's practices and I don't like that.
If that performance is true, then It means the clockspeed is well over 2GHz and for that reason I don't see how It can be under 50W, unless you didn't mean TBP but only TDP and even then I am a bit skeptical.
 
Last edited:

Glo.

Diamond Member
Apr 25, 2015
4,657
3,282
136
That would mean ~40% over RX 5500XT, which is pretty close to RTX 2060 6GB, but the amount of Vram is just not good and It copies Nvidia's practices and I don't like that.
If that performance is true, then It means the clockspeed is well over 2GHz and for that reason I don't see how It can be under 50W, unless you didn't mean TBP but only TDP and even then I am a bit skeptical.
And why would AMD care about that its "only 6 GB VRAM" since most likely 3050 is 4 GB VRAM GPU? This is the 149-169$ price tier we are talking about, after all.

When I said about 50W I talked about MOBILE SKU, not desktop, about which I talked with that quote of GTX 1660 Ti+15% performance. Desktop variant will have more power draw, obviously.
 

TESKATLIPOKA

Senior member
May 1, 2020
299
315
96
And why would AMD care about that its "only 6 GB VRAM" since most likely 3050 is 4 GB VRAM GPU? This is the 149-169$ price tier we are talking about, after all.
The last lowend(N14) from AMD had a 8GB version and this won't, although the 8GB version cost $199.
AMD may not care but for customers It could be a problem, but a low price could offset It.
We don't know how much Vram 3050 will have, If It will be only 4GB then N24 will have an advantage.

When I said about 50W I talked about MOBILE SKU, not desktop, about which I talked with that quote of GTX 1660 Ti+15% performance. Desktop variant will have more power draw, obviously.
So you actually meant mobile versions, that's realistic.
 
Last edited:

Glo.

Diamond Member
Apr 25, 2015
4,657
3,282
136
The last lowend(N14) from AMD had a 8GB version and this won't, although the 8GB version cost $199.
AMD may not care but for customers It could be a problem, but a low price could offset It.
We don't know how much Vram 3050 will have, If It will be only 4GB then N24 will have an advantage.


So you actually meant mobile versions, that's realistic.
And how much VRAM had 149$ GTX 1650 and 169$ RX 5500 XT? ;)

Those are the GPUs that puny little GPU will replace. Not the 8 GB one.
 

ASK THE COMMUNITY