Question Speculation: RDNA3 + CDNA2 Architectures Thread

[uzzi38]

Man I have been dying to make this one for a while now.

First rumours for RDNA3 are here so new thread time!

Just going to start off with this one for now: kopite7kimi on Twitter: "@VideoCardz Ah, I mean a simple mcm design with 10240 cores is not enough. Because the lift from RDNA2 to RDNA3 is much bigger than from RDNA1 to RDNA2. We should expect many big improvements in GFX11. 🤔" / Twitter

 

[PJVol]

60CUs Vs 70CUs is not that much different and if N32 clocks better performance could be quite close. We are talking a 15% delta so around 2.9Ghz for New Vs 2.5Ghz for the N31 design would give you equal compute, texture fillrate and RT performance.

What your assumption that n32 can be clocked higher than n31 at all (let alone +15%) is based upon?
Haven't heard anything to support it yet, and in fact I'm more inclined to believe the initial n32 revision is DOA.

However, the question remains of how much, if any, can be improved with the next revision, and whether it's still commercially viable.

 

[jpiniero]

At least at looking at 3dmark's results, the 7900 XTX can hit 2.9 Ghz just fine in a decent amount of samples... but there's a very high deviation. The 4080 and 4090 have a much tighter range.

 

[Timorous]

What your assumption that n32 can be clocked higher than n31 at all (let alone +15%) is based upon?
Haven't heard anything to support it yet, and in fact I'm more inclined to believe the initial n32 revision is DOA.

However, the question remains of how much, if any, can be improved with the next revision, and whether it's still commercially viable.

Smaller dies tend to clock higher in general. N22 clocks higher than N21 by 12.5% if you look at 6750XT vs 6950XT and 23.5% higher if you compare 6750XT vs 6800. If you want to look at original release the 6700XT has a 14.6% clock speed advantage vs the 6900XT

Then there is the fact the N31 seems to have been designed around a higher clock target that they cannot hit at sane power levels in games. If N32 goes some way to fix whatever causes that issue N32 will have a further improved v/f curve.

In addition 7900XTX has an 11% higher boost clock than the 6900XT (using this as neither are refresh parts)

Also checking the specs at Anandtech that W7800 config only has 128 rops so the delta is even less than I 1st though between that N31 config and the supposed N32 config.

So really given the fact smaller dies tend to clock better anyway and the supposed N31 clock vs power flaw and the N21 to N31 clock speed progression a 15% clock speed advantage for N32 vs N31 does not seem at all far fetched and is well within the. Just for reference an 11% increase in clock speed over the 6700XT would be 2.87Ghz which is about the same as a 15% hike from 2.5Ghz so there is that as well.

Given that it does not at all seem far fetched to think N32 can clock a bit higher than N31 and there is possibility to expect it could clock quite a lot higher. Obviously actual execution can be entirely different but ya know, speculation thread so that should always be a given.

 

[TESKATLIPOKA]

Not to be nitpicky, but the 7600S (28 CU) is a replacement for 6700S (28 CU) not 6800S (32 CU).
It has a TDP range of 50-75W. The 90W is already running the AMD equivalent of Nvidia dynamic boost. The 6800S has max TDP of 100W.
AMD has not changed much in N33 die with respect to N23 die. It's still 6nm, so the 25% improvement from just architecture change is ok (zen 2 to zen 3 on same node had 19% improvement). We are still getting the performance of the highest previous gen tier for this 'S' variant of GPU.

Not saying this is a great card, but it's not that bad a card to disparage as well. It's just..... Meh

There was no cut-down N22 in that test, so we compared it to 6800S and both of them consumed ~80W.

I seriously don't know from where you got this nonsense about 25% improvement from just architecture, there is clearly no such thing.

1080p	Cyberpunk 2077	Doom Eternal	F1 22	Far Cry 6	Ghostwire Tokyo	Guardians	Spider-Man	Average
RX 6800S	113%	118%	103%	105%	101%	103%	106%	107%
RX 7600S	100%	100%	100%	100%	100%	100%	100%	100%

RX 6800S has 14% more CU(Shader,TMU), but RX 7600S clocks higher.
What I found:
F1 22: 2350/2407MHz avg/max for 7600S at 90W. That should mean ~2300-2325MHz at the lowest.
F1 21: ~2400MHz is max frequency at 105W and then falls by 350MHz to 2050MHz at 80W, at 90W It looks to be ~2150-2175MHz. ComputerBase
So ~2300-2325Mhz vs ~2150-2175Mhz at comparable 90W, that's 7% difference.
My conclusion so far is that, If you compared the same configuration N33 vs N23 at comparable clocks, then performance would be just a bit better.

N33 looks like a FLOP. Why they even designed It is beyond me, at least If they used 5nm process, we would see some improvement in performance and better perf/W at lower TGP.

 

[PJVol]

At least at looking at 3dmark's results, the 7900 XTX can hit 2.9 Ghz just fine in a decent amount of samples... but there's a very high deviation

In TimeSpy the frequency is more like ~ specified "boost clock", i.e. 2.5 ghz.
3dmark results tell you nothing about power consumed, so 2.9 may be for heavily OC-ed/tuned/cooled GPU.

 

[Kepler_L2]

Smaller dies tend to clock higher in general. N22 clocks higher than N21 by 12.5% if you look at 6750XT vs 6950XT and 23.5% higher if you compare 6750XT vs 6800. If you want to look at original release the 6700XT has a 14.6% clock speed advantage vs the 6900XT

Then there is the fact the N31 seems to have been designed around a higher clock target that they cannot hit at sane power levels in games. If N32 goes some way to fix whatever causes that issue N32 will have a further improved v/f curve.

In addition 7900XTX has an 11% higher boost clock than the 6900XT (using this as neither are refresh parts)

Also checking the specs at Anandtech that W7800 config only has 128 rops so the delta is even less than I 1st though between that N31 config and the supposed N32 config.

So really given the fact smaller dies tend to clock better anyway and the supposed N31 clock vs power flaw and the N21 to N31 clock speed progression a 15% clock speed advantage for N32 vs N31 does not seem at all far fetched and is well within the. Just for reference an 11% increase in clock speed over the 6700XT would be 2.87Ghz which is about the same as a 15% hike from 2.5Ghz so there is that as well.

Given that it does not at all seem far fetched to think N32 can clock a bit higher than N31 and there is possibility to expect it could clock quite a lot higher. Obviously actual execution can be entirely different but ya know, speculation thread so that should always be a given.

W7800 has 160 ROPs.

 

[ryanjagtap]

There was no cut-down N22 in that test, so we compared it to 6800S and both of them consumed ~80W.

I seriously don't know from where you got this nonsense about 25% improvement from just architecture, there is clearly no such thing.

1080p	Cyberpunk 2077	Doom Eternal	F1 22	Far Cry 6	Ghostwire Tokyo	Guardians	Spider-Man	Average
RX 6800S	113%	118%	103%	105%	101%	103%	106%	107%
RX 7600S	100%	100%	100%	100%	100%	100%	100%	100%

RX 6800S has 14% more CU(Shader,TMU), but RX 7600S clocks higher.
What I found:
F1 22: 2350/2407MHz avg/max for 7600S at 90W. That should mean ~2300-2325MHz at the lowest.
F1 21: ~2400MHz is max frequency at 105W and then falls by 350MHz to 2050MHz at 80W, at 90W It looks to be ~2150-2175MHz. ComputerBase
So ~2300-2325Mhz vs ~2150-2175Mhz at comparable 90W, that's 7% difference.
My conclusion so far is that, If you compared the same configuration N33 vs N23 at comparable clocks, then performance would be just a bit better.

N33 looks like a FLOP. Why they even designed It is beyond me, at least If they used 5nm process, we would see some improvement in performance and better perf/W at lower TGP.

Yes, you are right. The 25% improvement estimate was too high. I read a few articles and see that the difference between the 6700S and 6800S is at the most 10-15%. So the new N33 7600S is not at all impressive.
Notebookcheck Comparison
Ultrabookreview Comparison

 

[jpiniero]

Also checking the specs at Anandtech that W7800 config only has 128 rops so the delta is even less than I 1st though between that N31 config and the supposed N32 config.

W7800 has to be N31 based.

 

[Timorous]

W7800 has 160 ROPs.

Anand have the spec wrong then which wouldn't be the 1st time. 128 doesn't make sense with 5SEs anyway unless you have 3 lots of 32 and 2 lots of 16 which is a really wonky config.

W7800 has to be N31 based.

I didn't say it wasn't.

I am saying a 7800XT based on that same config and a 7800XT based on full N32 with 15% higher boost clocks (~2.9Ghz) would have very similar performance.

 

[Timorous]

There was no cut-down N22 in that test, so we compared it to 6800S and both of them consumed ~80W.

I seriously don't know from where you got this nonsense about 25% improvement from just architecture, there is clearly no such thing.

1080p	Cyberpunk 2077	Doom Eternal	F1 22	Far Cry 6	Ghostwire Tokyo	Guardians	Spider-Man	Average
RX 6800S	113%	118%	103%	105%	101%	103%	106%	107%
RX 7600S	100%	100%	100%	100%	100%	100%	100%	100%

RX 6800S has 14% more CU(Shader,TMU), but RX 7600S clocks higher.
What I found:
F1 22: 2350/2407MHz avg/max for 7600S at 90W. That should mean ~2300-2325MHz at the lowest.
F1 21: ~2400MHz is max frequency at 105W and then falls by 350MHz to 2050MHz at 80W, at 90W It looks to be ~2150-2175MHz. ComputerBase
So ~2300-2325Mhz vs ~2150-2175Mhz at comparable 90W, that's 7% difference.
My conclusion so far is that, If you compared the same configuration N33 vs N23 at comparable clocks, then performance would be just a bit better.

N33 looks like a FLOP. Why they even designed It is beyond me, at least If they used 5nm process, we would see some improvement in performance and better perf/W at lower TGP.

Not a flop at all. Lower power consumption (good for laptops) drop in replaceable (allows aibs to re-use current designs) and slightly better performance in a package that costs AMD less to make so they can increase margin and lower the price for aibs at the same time.

Overall that sounds pretty good. Not as good as what was rumoured when people thought the dual issue designs was properly 2x the shaders but good none the less.

As for the desktop version give it fast gddr6 and it should hit close to 6700XT tier performance provided it clocks high enough.

Give AIBs the option to pair it with 16GB of ram as well and it could be a very very solid 1080p card an okay 1440p card.

 

[PJVol]

Smaller dies tend to clock higher in general. N22 clocks higher than N21 by 12.5% if you look at 6750XT vs 6950XT and 23.5% higher if you compare 6750XT vs 6800. If you want to look at original release the 6700XT has a 14.6% clock speed advantage vs the 6900XT

Then there is the fact the N31 seems to have been designed around a higher clock target that they cannot hit at sane power levels in games. If N32 goes some way to fix whatever causes that issue N32 will have a further improved v/f curve.

In addition 7900XTX has an 11% higher boost clock than the 6900XT (using this as neither are refresh parts)

Also checking the specs at Anandtech that W7800 config only has 128 rops so the delta is even less than I 1st though between that N31 config and the supposed N32 config.

So really given the fact smaller dies tend to clock better anyway and the supposed N31 clock vs power flaw and the N21 to N31 clock speed progression a 15% clock speed advantage for N32 vs N31 does not seem at all far fetched and is well within the. Just for reference an 11% increase in clock speed over the 6700XT would be 2.87Ghz which is about the same as a 15% hike from 2.5Ghz so there is that as well.

Given that it does not at all seem far fetched to think N32 can clock a bit higher than N31 and there is possibility to expect it could clock quite a lot higher. Obviously actual execution can be entirely different but ya know, speculation thread so that should always be a given

Wow, You're really not taking an easy way for speculating

What if, for the sake of clarity, we:

• agree that for now, the very existence of the rdna3 bugfix is nothing but wishful thinking.
• simplify the math and put aside rdna2 for a moment
• take the 71.4% of 7900XT (60CU/84CU) as the N32 performance baseline

Then if we agree to use correlated N22 => N21 scaling, then taking the actual clocks under gaming workload (not clocks from specs) an average clock delta for the N22 to N21 is around +8%, based on several "6000 refresh" review data.

So, we've got 71,4% + 8% = 77% of 7900XT performance, which depending on the review put it somewhere between 6800XT and 6900XT, closer to the former actually. And that's it )

 

[Ajay]

So, we've got 71,4% + 8% = 77% of 7900XT performance, which depending on the review put it somewhere between 6800XT and 6900XT, closer to the former actually. And that's it )

71.4 + 8.0 = 79.4 FIFY.

 

[AtenRa]

60CUs Vs 70CUs is not that much different and if N32 clocks better performance could be quite close. We are talking a 15% delta so around 2.9Ghz for New Vs 2.5Ghz for the N31 design would give you equal compute, texture fillrate and RT performance.

4 MCDs means both would have the same cache and memory bandwidth to go with that compute.

3 SEs Vs 5SEs means N31 version would have more rops and even with the clockspeed delta it would be ahead in this department but that would only show in cases where you were pixel fillrate limited.

So yea. Both would be pretty close in performance with just a 15% clockspeed advantage for N32 Vs that spec N31.

For raster perhaps but at higher consumption, for RT performance you cannot replace 10 RT cores with only 15% higher clocks.

Also, for this to work you will need a full NV32 die that will be able to clock at very high clocks and high consumption , instead of a cut off NV31 die with clocks that all dies can work.

We will find out in a few months.

 

[Kaluan]

From the recent potentially true MLID AMD APU info dump, Hawk Point potentially being a Phoneix+, Phoenix w/ RDNA3.5/RDNA3+ (same 6WGP config) in early 2024, along with virtually every other APU that year (including maybe the mainstream Zen5 IOD iGPU as well)... I see a tendency of AMD trying to get away from RDNA3 as fast as possible. Which feeds into the narrative that they borked up somehow the initial RDNA3 design indeed.

Will we see discrete RDNA3+ GPUs in 2024 as well?

 

[insertcarehere]

Overall that sounds pretty good. Not as good as what was rumoured when people thought the dual issue designs was properly 2x the shaders but good none the less.

As for the desktop version give it fast gddr6 and it should hit close to 6700XT tier performance provided it clocks high enough.

6800S (32 cu N23) at 80w is tested at ~7% faster than 7600S (28cu N33) at the same power envelope, but somehow give them both desktop-class memory and power, and Navi 33 would perform like a 6700xt while the former performs.. Like a 6600/50XT?

 

[PJVol]

71.4 + 8.0 = 79.4 FIFY.

I hope no one needs to explain the meaning of "%" symbol next to "8"

 

[Kepler_L2]

From the recent potentially true MLID AMD APU info dump, Hawk Point potentially being a Phoneix+, Phoenix w/ RDNA3.5/RDNA3+ (same 6WGP config) in early 2024, along with virtually every other APU that year (including maybe the mainstream Zen5 IOD iGPU as well)... I see a tendency of AMD trying to get away from RDNA3 as fast as possible. Which feeds into the narrative that they borked up somehow the initial RDNA3 design indeed.

Will we see discrete RDNA3+ GPUs in 2024 as well?

Nope, gfx11.5 is just for APUs.

 

[Ajay]

I hope no one needs to explain the meaning of "%" symbol next to "8"

Then 71.4 * 1.08. You wrote a math equation, so it's incorrect. Anyway....just being nitpicky.

 

[coercitiv]

6800S (32 cu N23) at 80w is tested at ~7% faster than 7600S (28cu N33) at the same power envelope, but somehow give them both desktop-class memory and power, and Navi 33 would perform like a 6700xt while the former performs.. Like a 6600/50XT?

I guess the rationale was that a 14% CU advantage resulted in a 7% performance advantage for N23 at ISO power. If the desktop N33 brings CU count to parity and scales clocks better by something like ~10%, then it could lead the 6600XT by 15%+ at 1080p. This happens to be the rought performance delta between 6600XT and 6700XT at 1080p.

That being said, I would not consider this 6700XT tier, resolution scaling will be problematic. Performance gap between 6600XT and 6700XT grows to 23% at 1440p.

 

[Timorous]

Wow, You're really not taking an easy way for speculating

What if, for the sake of clarity, we:

• agree that for now, the very existence of the rdna3 bugfix is nothing but wishful thinking.
• simplify the math and put aside rdna2 for a moment
• take the 71.4% of 7900XT (60CU/84CU) as the N32 performance baseline

Then if we agree to use correlated N22 => N21 scaling, then taking the actual clocks under gaming workload (not clocks from specs) an average clock delta for the N22 to N21 is around +8%, based on several "6000 refresh" review data.

So, we've got 71,4% + 8% = 77% of 7900XT performance, which depending on the review put it somewhere between 6800XT and 6900XT, closer to the former actually. And that's it )

Not sure this is easier given the number of baked in assumptions.

It seems 7900XT and XTX clock about the same in games on average and scaling appears to match the relative compute performance fairly closely.

as such 71.4% with a 15% clock bump would be around 82% of 7900XT and 70/84 is 83%.

So a full N32 design that clocks around 2.9Ghz in games would roughly match a 70CU N31 design that clocks around 2.5Ghz in games and both would be in 6950XT ballpark territory.

6800S (32 cu N23) at 80w is tested at ~7% faster than 7600S (28cu N33) at the same power envelope, but somehow give them both desktop-class memory and power, and Navi 33 would perform like a 6700xt while the former performs.. Like a 6600/50XT?

An OCd 6650XT is about 10% ahead of the stock 6650XT. Allow that to represent the likely higher core clock and faster VRAM of a 7600XT and then factor in the approx 9% IPC increase and you get a 7600XT that is about 20% faster than a stock 6650XT and at 1440p that would be around 6700XT performance on average.

 

[TESKATLIPOKA]

Not a flop at all. Lower power consumption (good for laptops) drop in replaceable (allows aibs to re-use current designs) and slightly better performance in a package that costs AMD less to make so they can increase margin and lower the price for aibs at the same time.

Overall that sounds pretty good. Not as good as what was rumoured when people thought the dual issue designs was properly 2x the shaders but good none the less.

As for the desktop version give it fast gddr6 and it should hit close to 6700XT tier performance provided it clocks high enough.

Give AIBs the option to pair it with 16GB of ram as well and it could be a very very solid 1080p card an okay 1440p card.

A bit higher performance a bit better power consumption, the same Vram and ~$10-15 saved on producing these chips compared to N23. This sounds pretty good to you? Maybe If there was no competition.

What advantage does N33 have over RTX 4060 laptop? Nothing except lower production cost. AMD will need to lower price as much as possible for AIBs to sell something.

 

[Kaluan]

Nope, gfx11.5 is just for APUs.

Well, that's unfortunate. I guess the alleged super-APU "Sarlak" will be RDNA3+ fastest incarnation then.
Will defo look into it if I have the cash on hand, will probably make for a helluva mini/micro desktop PC/gaming console!

A bit higher performance a bit better power consumption, the same Vram and ~$10-15 saved on producing these chips compared to N23. This sounds pretty good to you? Maybe If there was no competition.

What advantage does N33 have over RTX 4060 laptop? Nothing except lower production cost. AMD will need to lower price as much as possible for AIBs to sell something.

There, that's the pretty good part. How exactly is that a bad thing and not a win for us?
If anything, have people not learned anything from the success of Polaris?

 

[PJVol]

Not sure this is easier given the number of baked in assumptions.

It's funny that the only assumption in my post was in fact made by you (and which I agreed with) that the N32 clocks margin resulted from overall Cac reduction, at the very least is similar to the N22.
The rest is math.

 

[insertcarehere]

An OCd 6650XT is about 10% ahead of the stock 6650XT. Allow that to represent the likely higher core clock and faster VRAM of a 7600XT and then factor in the approx 9% IPC increase and you get a 7600XT that is about 20% faster than a stock 6650XT and at 1440p that would be around 6700XT performance on average.

- The 7600S that was tested was clocking at ~2.4 ghz, we don't know what the 6800S clocked at in the comparison, but from other tests on the laptop it tends to settle at around ~2.1ghz in an 80w envelope, so 7600S most likely enjoyed a sizable clock speed advantage...and still lost out.
- The 9% IPC increase was based on comparing a 6900XT vs 7900XT at 4K, but there are other factors at play there which I have already explained before:

To be frank I don't think we can conclude that dual issue brings performance improvement by itself. Yes the computerbase test comparing 7900XT vs 6900XT shows some improvement (9% average at 4k) but those are not equal GPUs even if you normalize CU/clocks.
- 7900XT has 800gb/s of vram bandwidth vs 512 gb/s for 6900XT, this is counteracted somewhat by the higher IC the latter has, but at 4k the 7900XT still should have significantly more usable bandwidth.
- 7900XT has 192 ROPs vs 128 ROPs for 6900 XT, so it has 50% higher Pixel Rate even at the same clocks.

How much of the observed improvement is due to 2x FP32 as opposed to the 7900XT just being more endowed in bandwidth and GPU front end?

A 7600XT wouldn't have 50% more bandwidth nor 50% more ROPs than a 6650XT...

 

[TESKATLIPOKA]

There, that's the pretty good part. How exactly is that a bad thing and not a win for us?
If anything, have people not learned anything from the success of Polaris?

Let's say 7600M XT laptop will be sold $50 cheaper than a comparable one with RTX 4060. Is It worth It?
At $1149 vs $1199 It's only 4.2% cheaper
At $1449 vs $1499 It's only 3.34% cheaper
Would you buy It over Nvidia? I personally wouldn't.

At least 7600M(S) against RTX 4050 laptop looks pretty decent, but the question is If It is worth It compared to the stronger GPUs like RXT 4060 or RX 7600M XT.

Success of Polaris? Are you sure? AMD had to lower prices so much that it barely made anything per card. From when is this considered success? Not to mention that in mobile they had almost zero presence because they were power hungry compared to competition.

edit:
Not sure if in desktop 7600XT will do that much better.
Here, a difference of $50 would be more significant, true.
On the other hand, 4060(full Ad107) will be more efficient, and I think will have better OC potential.
They should release the cards finally.