• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

[Computerbase] RDNA 2 CU Scaling

ChiefDada

Gold Member
Preface: THIS THREAD IS FOR THOSE INTERESTED IN TECH ANALYSIS/DISCUSSION OF RDNA 2 AND THE DESIGN CHOICES FOR PS5 AND SERIES X. NO CONSOLE WARS PLEASE!

https://www.computerbase.de/2021-03/amd-radeon-rdna2-rdna-gcn-ipc-cu-vergleich/2/

Awesome analysis that I think is the most comprehensive in comparing RDNA 2 CU scalability between AMD 6000 series cards, all fixed at 2ghz clock frequencies. Results posted below, but I wanted to point out a crucial point from a different but related test they conducted that determined RDNA 1 CUs are actually faster than RDNA 2 CU due to the shorter ALU pipeline (this should serve as a reminder to some that RDNA 2 isn't inherently better than RDNA 1 across the board AMD) .

RDNA 2 clocks significantly higher than RDNA and achieves a not insignificant part of its additional performance from it. In order to reach the higher frequencies, the pipelines within the ALUs were lengthened, among other things, which increases the latencies and thus reduces the computing power per CU with RDNA 2 compared to RDNA, despite actually having the same capabilities for latency-sensitive calculations.

RDNA 2 has slowed down a bit​

You can see that in the tests. RDNA 2 is on average 4 and 5 percent slower than RDNA with the same computing power. At the same time, this also shows that the longer pipelines for RDNA 2 were a good decision. In practice, this 4 percent reduction in computing power is offset by a clock rate that is around 24 percent higher for Navi 21 (RX 6800 XT) and a clock rate that is around 40 percent higher for Navi 22 (RX 6700 XT) - a good exchange: IPC for frequency.

Applying the findings below to the premium consoles, there are a few interesting facts from my perspective:

1. Cerny's argument of CU utilization doesn't pan out in the games tested below. CU scalability remains relatively constant from 1080p-4k. Although it is possible that current gen games could yield a different outcome.

2. Series consoles chose RDNA 2 design with inherent CU latency increase compared to RDNA 1, without offsetting via higher clock frequency (2ghz+) as AMD and Sony have done with 6000 series and PS5, respectively. Based on the testing, it doesn't appear as though Series X 560gb/s throughput would be enough to compensate to achieve the levels AMD was able to with faster clocks/cache. As shown below, A 50% CU advantage AND 33% bandwidth advantage for the 6800 over the 6700xt resulted in only 36% performance increase at 4k (again, both GPU clock frequencies fixed at 2ghz). Series X has 44% CU advantage and 25% bandwidth advantage ceiling over PS5 with its fastest memory segment. If anyone has information as to why Microsoft deviated from AMD RDNA 2 strategy, I would be interested to learn about this (proprietery hw/sw, etc.).


Data Parameters

doursXZ.jpg





1080p Performance
L5MdeBz.jpg


1440p Performance
rd2D46l.jpg


4k Performance

rlLOITq.jpg
 

Pagusas

Elden Member
whats the thermal and power difference between less CU but higher frequency, and more CU but lower frequency? I could be mistaken, but I thought the reason MS choose Fat, wide and slower (more CU's, Memory bandwith, but slower frequency) was for thermal and energy sake, where as Sony took the hotter/more energy required state. I wonder too if it was easier to create a Series S and Series X via CU scaling.
 
Last edited:

01011001

Banned
Microsoft most likely went that route due to heat. they can have a more compact cooling solution and one that uses standard materials.

also die SX uses usually less power in like for like scenarios, which might have been another factor they were looking for... didn't they get into issues with the power consumption on the Xbox One due to the fast boot being the default setting? mabe this time they were more conscious about that in general just to be save
 
Last edited:

assurdum

Banned
Preface: THIS THREAD IS FOR THOSE INTERESTED IN TECH ANALYSIS/DISCUSSION OF RDNA 2 AND THE DESIGN CHOICES FOR PS5 AND SERIES X. NO CONSOLE WARS PLEASE!

https://www.computerbase.de/2021-03/amd-radeon-rdna2-rdna-gcn-ipc-cu-vergleich/2/

Awesome analysis that I think is the most comprehensive in comparing RDNA 2 CU scalability between AMD 6000 series cards, all fixed at 2ghz clock frequencies. Results posted below, but I wanted to point out a crucial point from a different but related test they conducted that determined RDNA 1 CUs are actually faster than RDNA 2 CU due to the shorter ALU pipeline (this should serve as a reminder to some that RDNA 2 isn't inherently better than RDNA 1 across the board AMD) .



Applying the findings below to the premium consoles, there are a few interesting facts from my perspective:

1. Cerny's argument of CU utilization doesn't pan out in the games tested below. CU scalability remains relatively constant from 1080p-4k. Although it is possible that current gen games could yield a different outcome.

2. Series consoles chose RDNA 2 design with inherent CU latency increase compared to RDNA 1, without offsetting via higher clock frequency (2ghz+) as AMD and Sony have done with 6000 series and PS5, respectively. Based on the testing, it doesn't appear as though Series X 560gb/s throughput would be enough to compensate to achieve the levels AMD was able to with faster clocks/cache. As shown below, A 50% CU advantage AND 33% bandwidth advantage for the 6800 over the 6700xt resulted in only 36% performance increase at 4k (again, both GPU clock frequencies fixed at 2ghz). Series X has 44% CU advantage and 25% bandwidth advantage ceiling over PS5 with its fastest memory segment. If anyone has information as to why Microsoft deviated from AMD RDNA 2 strategy, I would be interested to learn about this (proprietery hw/sw, etc.).


Data Parameters

doursXZ.jpg





1080p Performance
L5MdeBz.jpg


1440p Performance
rd2D46l.jpg


4k Performance

rlLOITq.jpg
But all the gpu use the same frequency. Of course the Cerny argumentation not pan out there.
 
Last edited:

assurdum

Banned
Sure but you can still analyze in a vacuum. Some games can utilize high core count better or more poorly than others.
Of course. But Cerny has never said less CU at the same frequency work better than more CUs. He talked to use less CU at higher frequency compared more CUs with lower frequency. It's a complete different argument. Neither he has said it was the best approach possible in absolute terms, to be fair.
 
Last edited:

Riky

$MSFT
The clock speed for the Series X was set to meet the target of twice the Xbox One X power, 12tflops. That was the target, the clock speed is set exactly for that, it could probably clock higher but they didn't feel the need to as there is also a 25% uplift in efficiency on top per CU over last gen. Size, heat and noise also come onto this.
Jason Ronald has talked about this several times.
 
Last edited:

assurdum

Banned
The clock speed for the Series X was set to meet the target of twive the Xbox One X power, 12tflops. That was the target, the clock speed is set exactly for that, it could probably clock higher but they didn't feel the need to as there is also a 25% uplift in efficiency on top per CU over last gen. Size, heat and noise also come onto this.
Jason Ronald has talked about this several times.
annoyed for gods sake GIF
 

ChiefDada

Gold Member
whats the thermal and power difference between less CU but higher frequency, and more CU but lower frequency? I could be mistaken, but I thought the reason MS choose Fat, wide and slower (more CU's, Memory bandwith, but slower frequency) was for thermal and energy sake, where as Sony took the hotter/more energy required state. I wonder too if it was easier to create a Series S and Series X via CU scaling.

Yes but I think performance per watt is the focus for AMD. Whether increasing GPU frequency or increasing memory bandwidth (256 vs 320bit) will result in higher power consumption. AMD saw that gpu frequency increase was better route from a performance/watt standpoint because infinity cache with effective bw multiplier was performant enough to more than counterbalance potential limitations from 256bit memory. At least that's my understanding.

Of course. But Cerny has never said if you less CU at the same frequency they work better. He talked to use less CU at higher frequency. It's a complete different argument.

I agree and never intended to suggest that's what he said. I'm just saying he evoked a known issue of high CU utilization for certain workloads which is valid. Even Computerbase referenced this issue (see below). All I am saying is that in this particular test, with these particular games, CU utilization and scaling was not an issue.

Navi 22 on the Radeon RX 6700 XT has over 40 compute units and thus 2,560 shader units, while Navi 21 on the Radeon RX 6900 XT has over 80 CUs and 5,120 ALUs. In the data sheet, the increase in performance is huge because the clock does not drop significantly. In practice, however, twice the number of units must first be utilized, which has not always worked properly in the past.


No one can blame me for not trying:messenger_winking:
 

ChiefDada

Gold Member
The clock speed for the Series X was set to meet the target of twive the Xbox One X power, 12tflops. That was the target, the clock speed is set exactly for that, it could probably clock higher but they didn't feel the need to as there is also a 25% uplift in efficiency on top per CU over last gen. Size, heat and noise also come onto this.
Jason Ronald has talked about this several times.

Did you read the analysis? If what you're saying were true then that would be short-sided. In order for the higher latency of RDNA 2 CU to be mitigated and actually make sense in the long run, then clock frequencies need to be INCREASED at 2gz+ speed as AMD and PS5 have done.
 

Riky

$MSFT
Did you read the analysis? If what you're saying were true then that would be short-sided. In order for the higher latency of RDNA 2 CU to be mitigated and actually make sense in the long run, then clock frequencies need to be INCREASED at 2gz+ speed as AMD and PS5 have done.

Xbox One X wasn't RDNA, so no. They have stated many times they were looking for 2x the performance of Xbox One X, they achieved that.

The closest RDNA2 card at launch of the range was the 6800 and the game clock is almost identical to Series X.
 
Last edited:

Lysandros

Member
Preface: THIS THREAD IS FOR THOSE INTERESTED IN TECH ANALYSIS/DISCUSSION OF RDNA 2 AND THE DESIGN CHOICES FOR PS5 AND SERIES X. NO CONSOLE WARS PLEASE!

https://www.computerbase.de/2021-03/amd-radeon-rdna2-rdna-gcn-ipc-cu-vergleich/2/

Awesome analysis that I think is the most comprehensive in comparing RDNA 2 CU scalability between AMD 6000 series cards, all fixed at 2ghz clock frequencies. Results posted below, but I wanted to point out a crucial point from a different but related test they conducted that determined RDNA 1 CUs are actually faster than RDNA 2 CU due to the shorter ALU pipeline (this should serve as a reminder to some that RDNA 2 isn't inherently better than RDNA 1 across the board AMD) .



Applying the findings below to the premium consoles, there are a few interesting facts from my perspective:

1. Cerny's argument of CU utilization doesn't pan out in the games tested below. CU scalability remains relatively constant from 1080p-4k. Although it is possible that current gen games could yield a different outcome.
Thanks for sharing. First, i would like to remind you of Cerny's exact quote from Road to PS5 "It is easier to fully use 36 CUs in parallel than to use 48 CUs, when the triangles are small it is much harder to fill all those CUs with useful work." meaning higher the polycount/geometry complexity (as expected from the 'real' current generation) is greater the difficulty/efficiency loss.

Second, the ROP architectures of RDNA 1 and 2 are different. RDNA 1 desktop GPU's (and PS5 as an exception) have twice the number of z-stencil/depth ROP hardware compared to RDNA 2 and XSX. Thus RDNA 1 being ~4-5% faster at the same clock could also be related to this besides the shorter ALU pipeline.
 
Last edited:

Vognerful

Member
Is it possible in any shape or form to boost the SoCs of these consoles beyond the current frequency? I understand for example that the PS5 has a power limit or load, and I think a fixed frequency on Xbox.
 
Something else to keep in mind is it's not just about the raw hardware setups i.e having more CUs or higher peak bandwidth; it's about hardware customizations and the API libraries especially.

Knowing now about the aspects of latency increasing between RDNA1 to RDNA2 due to a longer pipeline, it puts curiosity both on MS making the Series X Shader Arrays larger than the usual for RDNA2 (7 Dual CUs per Shader Array in MS's case for Series X) and Sony adding cache scrubbers to PS5 (probably to help mitigate the inherent added latency in RDNA2 as well as cut down on bus accesses via selective cache line flushes instead of flushing out the entire cache for each modify).

In MS's case I think their choices were really down to ensuring easy scalability between the S and X, and enabling good BC as well as having a design in Series X that would be relatively even with PS5 and also scalable for Azure. I'm stressing "even" because between the real-world performance results in bigger RDNA2 GPUs having smaller gains over weaker RDNA2 GPUs even while having larger gaps (on paper) than XSX does over PS5 in arithmetic, combined with differences in API libraries between Sony & Microsoft, and various customizations on both platforms, the theoretical gap is never going to manifest in a way where Series X is just "pulling well ahead" over PS5, that's for the entirety of the generation.

You'll get some cases of multiplats having slight edges on Series X over PS5, and cases of some having advantages on PS5 over Series X, but there's never going to be a point where Series X is just pulling ahead consistently in all multiplat comparisons and earlier results should've indicated that. Any sheer notable differences in the respective designs will be shown through exclusives (1P, 3P with heavy 1P involvement) and in terms of general technical prowess Sony's teams still have the advantage and I'd argue they have the artistic advantage as well more cases than not, outside of maybe gritty apocalyptic military scifi/horror where I think you can say The Coalition are pretty high up there (even there though, depending on your classification you can fit Naughty Dog in that category and they end up taking it).

The only other area I can say where a MS team might have a leg-up on a Sony team in terms of at least technical ability is racing because extreme LOD pop-in aside, the new Forzatech engine is pushing more than Polyphony's in GT7 by a decent margin. Polyphony are gonna have to step up the features of their engine or better yet just fork it into Decima for GT8.

So basically, if you want the best of 3P, go with the console with the ecosystem or features you prefer, because actual performance is gonna be a wash/virtually identical 90% of the time. If you want the best 1P and exclusives, well one platform comes from a lineage that has proven it can do those things many times over, the other comes from a lineage where their proving of that has been very scattershot at best and only for very select genres. That could drastically change in the future but talking long-term that's what the results have generally been.

Still though, if there's just that very specific game you want to play that's on one platform and not the other, and it means that much to you, you know which way to go.
 

onesvenus

Member
Something else to keep in mind is it's not just about the raw hardware setups i.e having more CUs or higher peak bandwidth; it's about hardware customizations and the API libraries especially.

Knowing now about the aspects of latency increasing between RDNA1 to RDNA2 due to a longer pipeline, it puts curiosity both on MS making the Series X Shader Arrays larger than the usual for RDNA2 (7 Dual CUs per Shader Array in MS's case for Series X) and Sony adding cache scrubbers to PS5 (probably to help mitigate the inherent added latency in RDNA2 as well as cut down on bus accesses via selective cache line flushes instead of flushing out the entire cache for each modify).

In MS's case I think their choices were really down to ensuring easy scalability between the S and X, and enabling good BC as well as having a design in Series X that would be relatively even with PS5 and also scalable for Azure. I'm stressing "even" because between the real-world performance results in bigger RDNA2 GPUs having smaller gains over weaker RDNA2 GPUs even while having larger gaps (on paper) than XSX does over PS5 in arithmetic, combined with differences in API libraries between Sony & Microsoft, and various customizations on both platforms, the theoretical gap is never going to manifest in a way where Series X is just "pulling well ahead" over PS5, that's for the entirety of the generation.

You'll get some cases of multiplats having slight edges on Series X over PS5, and cases of some having advantages on PS5 over Series X, but there's never going to be a point where Series X is just pulling ahead consistently in all multiplat comparisons and earlier results should've indicated that. Any sheer notable differences in the respective designs will be shown through exclusives (1P, 3P with heavy 1P involvement) and in terms of general technical prowess Sony's teams still have the advantage and I'd argue they have the artistic advantage as well more cases than not, outside of maybe gritty apocalyptic military scifi/horror where I think you can say The Coalition are pretty high up there (even there though, depending on your classification you can fit Naughty Dog in that category and they end up taking it).

The only other area I can say where a MS team might have a leg-up on a Sony team in terms of at least technical ability is racing because extreme LOD pop-in aside, the new Forzatech engine is pushing more than Polyphony's in GT7 by a decent margin. Polyphony are gonna have to step up the features of their engine or better yet just fork it into Decima for GT8.

So basically, if you want the best of 3P, go with the console with the ecosystem or features you prefer, because actual performance is gonna be a wash/virtually identical 90% of the time. If you want the best 1P and exclusives, well one platform comes from a lineage that has proven it can do those things many times over, the other comes from a lineage where their proving of that has been very scattershot at best and only for very select genres. That could drastically change in the future but talking long-term that's what the results have generally been.

Still though, if there's just that very specific game you want to play that's on one platform and not the other, and it means that much to you, you know which way to go.

NO CONSOLE WARS PLEASE!
 

Lysandros

Member
Did you read the analysis? If what you're saying were true then that would be short-sided. In order for the higher latency of RDNA 2 CU to be mitigated and actually make sense in the long run, then clock frequencies need to be INCREASED at 2gz+ speed as AMD and PS5 have done.
Well with the lower frequency/more CUs approach Microsoft reached their initial goal of 12 TF with the added benefits of having a smaller and more power efficient machine at the cost of loosing some GPU fixed function throughput and performance efficiency. This doesn't seem to be a particularly risky trade-off to me. XSX stays highly competitive against PS5, not that it is particularly underperforming or anything.
 
onesvenus onesvenus I don't get it; how was I console warring? Don't think I said anything ridiculous tbh.

MS opted for this solution (lower clocks), so there would be less broken chips. It was in some of their presentation.

I don't think 'broken chips' was ever a reason why, considering what limits RDNA2 can be pushed frequency-wise. MS just wanted an approach that was more manageable for their cooling design targets, and something that would be manageable for Series X usage in server clusters via Azure (where generally, going wider but slower wins out).

The only 'source' that ever claimed something about high PS5 APU failure rates was Bloomberg, and Sony denied that report themselves. No one else has come out with similar reports...well no one credible, anyway.
 
Last edited:

M1chl

Currently Gif and Meme Champion
onesvenus onesvenus I don't get it; how was I console warring? Don't think I said anything ridiculous tbh.



I don't think 'broken chips' was ever a reason why, considering what limits RDNA2 can be pushed frequency-wise. MS just wanted an approach that was more manageable for their cooling design targets, and something that would be manageable for Series X usage in server clusters via Azure (where generally, going wider but slower wins out).

The only 'source' that ever claimed something about high PS5 APU failure rates was Bloomberg, and Sony denied that report themselves. No one else has come out with similar reports...well no one credible, anyway.
Obviously bigger chips on higher frequency have way bigger chance of failure than smaller chip, like it's inside PS5. And it's not dis of PS5, it's contrary. I sometimes wonder how much chips for 3080/3090 ends up in bin. After all it 3080 have almost 30 billion transistors, it's crazy that they even make it work...
 

ChiefDada

Gold Member
Well with the lower frequency/more CUs approach Microsoft reached their initial goal of 12 TF with the added benefits of having a smaller and more power efficient machine at the cost of loosing some GPU fixed function throughput and performance efficiency.

The bolded is the part I don't understand. Why would your design goals be so bound by an arbitrary number that is made even more meaningless between vastly different architectures? That being said, if form factor was the main goal then I say mission accomplished.

onesvenus onesvenus onesvenus onesvenus I don't get it; how was I console warring? Don't think I said anything ridiculous tbh.

I am equally confused as I thought your commentary was fair and well thought out.
 
Obviously bigger chips on higher frequency have way bigger chance of failure than smaller chip, like it's inside PS5. And it's not dis of PS5, it's contrary. I sometimes wonder how much chips for 3080/3090 ends up in bin. After all it 3080 have almost 30 billion transistors, it's crazy that they even make it work...

That's true, but even in the case of that Series X is kind of an outlier because there are RDNA2 GPUs much bigger than it that can also clock higher, so it seems MS's reasoning was a deliberate choice not just for their console footprint design & cooling budget, but also because they wanted something that could be cheap but powerful enough to run in an Azure server cluster without pushing the energy bill too high.

Plus, again, they wanted something fit for BC and running multiple Series S simulated instances. They needed a wider design for that at scale.

The bolded is the part I don't understand. Why would your design goals be so bound by an arbitrary number that is made even more meaningless between vastly different architectures? That being said, if form factor was the main goal then I say mission accomplished.



I am equally confused as I thought your commentary was fair and well thought out.

Guess some people just want to hear about dat "power delta" finally manifesting as if by magic any time now. But those same people only look at TF and raw peak RAM bandwidth, and don't usually understand intricacies in other parts of system design like cache scrubbing, bus latency, cache coherency, pixel fillrate (< all particular features of PS5), virtualized RAM partition, mip-blending (< all particular features of Series X) etc.
 

DaGwaphics

Member
whats the thermal and power difference between less CU but higher frequency, and more CU but lower frequency? I could be mistaken, but I thought the reason MS choose Fat, wide and slower (more CU's, Memory bandwith, but slower frequency) was for thermal and energy sake, where as Sony took the hotter/more energy required state. I wonder too if it was easier to create a Series S and Series X via CU scaling.

Very true. You can't even have the discussion without figuring those areas in. XSX chip seems perfectly suited for the goal of building a console the current size. Plus, it can run 4 X1 games independently, that was probably a design consideration as well.

Edit: Should have read the whole thread, @ thicc_girls_are_teh_best thicc_girls_are_teh_best beat me to it. Although, I don't think the XSX chip is capable of running multiple XSS instances, major CPU issues would creep up there since the XSS CPU is almost as strong.
 
Last edited:

01011001

Banned
Very true. You can't even have the discussion without figuring those areas in. XSX chip seems perfectly suited for the goal of building a console the current size. Plus, it can run 4 X1 games independently, that was probably a design consideration as well.

backwards compatibility was most likely a bigger influence on both consoles than most people think
the PS5's CU count is definitely a conscious choice for example. I don't think there was ever a chance that they would not have 36CUs after they comminted to full backwards compatibility
 
Very true. You can't even have the discussion without figuring those areas in. XSX chip seems perfectly suited for the goal of building a console the current size. Plus, it can run 4 X1 games independently, that was probably a design consideration as well.

Edit: Should have read the whole thread, @ thicc_girls_are_teh_best thicc_girls_are_teh_best beat me to it. Although, I don't think the XSX chip is capable of running multiple XSS instances, major CPU issues would creep up there since the XSS CPU is almost as strong.

Actually that's a good point, I forgot about the CPU in that equation. Maybe games very light on CPU usage WRT cores & threads could work, but that would be a small portion of total games.

Otherwise, yeah, mainly One S simulated instances would be the main usage, with some work o boost resolution and framerate where possible.
 

onesvenus

Member
onesvenus onesvenus I don't get it; how was I console warring? Don't think I said anything ridiculous tbh.

I am equally confused as I thought your commentary was fair and well thought out.

All is fair when you base your ideas around facts. When you say subjective things like
in terms of general technical prowess Sony's teams still have the advantage and I'd argue they have the artistic advantage as well more cases than not, outside of maybe gritty apocalyptic military scifi/horror where I think you can say The Coalition are pretty high up there (even there though, depending on your classification you can fit Naughty Dog in that category and they end up taking it).
or
The only other area I can say where a MS team might have a leg-up on a Sony team in terms of at least technical ability is racing because extreme LOD pop-in aside, the new Forzatech engine is pushing more than Polyphony's in GT7 by a decent margin. Polyphony are gonna have to step up the features of their engine or better yet just fork it into Decima for GT8.
You are only inviting to console warring.

I could have just laughed at your post and namedrop iD for example but that would have only lead to console warring
 

winjer

Gold Member
Obviously bigger chips on higher frequency have way bigger chance of failure than smaller chip, like it's inside PS5. And it's not dis of PS5, it's contrary. I sometimes wonder how much chips for 3080/3090 ends up in bin. After all it 3080 have almost 30 billion transistors, it's crazy that they even make it work...

In N7 the reported number, by TSMC, for failure rate is of 0.09 sq/cm.

The SoC for the PS5 should get around 76% yield rate. For the Series X, it should be around 73%
The die for the GA102-350-A1 would get around 58%.
But unlike consoles, many dies with defects can be cutdown and reused to make lower skews, for the 3090/3080 family.

EDIT: I forgot one detail. THE PS5 and Series X have a few CUs that are disabled, to improve yield rate. So these rates should be a bit higher.
 
Last edited:

Xyphie

Member
But unlike consoles, many dies with defects can be cutdown and reused to make lower skews, for the 3090/3080 family.

They do have fault tolerance in the GPU because they don't use the full thing (36/40 and 52/56, so one WGP per shader engine can fail). Likewise with caches and such which will have redundancy built-in.
 

Dream-Knife

Banned
Xbox One X wasn't RDNA, so no. They have stated many times they were looking for 2x the performance of Xbox One X, they achieved that.

The closest RDNA2 card at launch of the range was the 6800 and the game clock is almost identical to Series X.
6800 runs at 2170 in game.

In before someone says PS5 is RDNA 1.5.
Primitive shaders vs mesh shaders.
 
Last edited:

Riky

$MSFT
Card never goes that low unless you're idle.

These cards are also incredibly easy to OC. 2400+ is quite achievable.

Maybe so, but that's what AMD call it, also MS talked a lot about sustained performance, so a set clock speed was obviously a goal.
The advantages of the larger GPU also come in shader units, TMUs and texture rate.

Also the RDNA2 render backend is different which leads to hardware VRS being possible on Series consoles.
Series consoles also have additional customisations outside of RDNA2 with core adjustments for ML and filters for SFS.
It's quite a custom part.
 

Dream-Knife

Banned
Maybe so, but that's what AMD call it, also MS talked a lot about sustained performance, so a set clock speed was obviously a goal.
The advantages of the larger GPU also come in shader units, TMUs and texture rate.

Also the RDNA2 render backend is different which leads to hardware VRS being possible on Series consoles.
Series consoles also have additional customisations outside of RDNA2 with core adjustments for ML and filters for SFS.
It's quite a custom part.
I wasn't arguing with you, just telling you that series x isn't close to 6800 clock speeds.

 

Riky

$MSFT
I wasn't arguing with you, just telling you that series x isn't close to 6800 clock speeds.

The base or game clock is and that's the nearest GPU to Series X, like I said MS wanted sustained performance.
 
As a stupid in this topic, I have 4 takeaways.

1. PS5: Narrow but faster.
2. XboxSX: wider but slower.

3. Both consoles are virtually identical.

4. The difference is made by the devs.
I wouldn't say virtually identical.

Those are decisions that can lead to vast different results.

In a way PS5 was designed to mitigate the downfalls of narrow but fast and seems to have done it's job well. When it comes to running code on it, last gen code that is, it might even be easier to make it perform better, as it doesn't need to be as parallelized, theoretically.

Xbox Series S/X is theoretically the most forward looking, and that might give it an edge towards the end of the generation. But it also might be a non-factor. Depends on how hardware (and code) evolves around these consoles.
In N7 the reported number, by TSMC, for failure rate is of 0.09 sq/cm.

The SoC for the PS5 should get around 76% yield rate. For the Series X, it should be around 73%

EDIT: I forgot one detail. THE PS5 and Series X have a few CUs that are disabled, to improve yield rate. So these rates should be a bit higher.
Yes.

And lower clocks with less aggressive turbo will also make a lot of chips more passable for Microsoft. I believe Series X yields probably benefit from the strategy, as Sony will get some semi-good processors, just not good enough for the load they will put them through. Liquid metal probably helps a lot though.

This said Sony is saving precious yield space with smaller dies so in the end it might just work out when it comes to cost.
 
Last edited:

Riky

$MSFT
The base or game clocks don't really exist on modern gpus.

A bone stock 6800 in game runs at 2225 as shown in that link.
Nobody denys that, it's just the closest GPU to Series X which is at the end of the day is an APU and has much different considerations, my point was the game clock as AMD call it is actually slightly below the Series X, no debate it can go higher at all.
 

assurdum

Banned
Did you read the analysis? If what you're saying were true then that would be short-sided. In order for the higher latency of RDNA 2 CU to be mitigated and actually make sense in the long run, then clock frequencies need to be INCREASED at 2gz+ speed as AMD and PS5 have done.
Don't waste your time. He's not interested in anything tech wise which it's not celebrating the XBX power. He parrot the MS PR marketing and nothing more in his replies.
 
Last edited:

assurdum

Banned
I wouldn't say virtually identical.

Those are decisions that can lead to vast different results.

In a way PS5 was designed to mitigate the downfalls of narrow but fast and seems to have done it's job well. When it comes to running code on it, last gen code that is, it might even be easier to make it perform better, as it doesn't need to be as parallelized, theoretically.

Xbox Series S/X is theoretically the most forward looking, and that might give it an edge towards the end of the generation. But it also might be a non-factor. Depends on how hardware (and code) evolves around these consoles.

Yes.

And lower clocks with less aggressive turbo will also make a lot of chips more passable for Microsoft. I believe Series X yields probably benefit from the strategy, as Sony will get some semi-good processors, just not good enough for the load they will put them through. Liquid metal probably helps a lot though.

This said Sony is saving precious yield space with smaller dies so in the end it might just work out when it comes to cost.
There is nothing in the XSX yield can be pushed more compared the ps5 for the end of generation. What you see until now in multiplat you will get again in the same size for the end of the generation. Not sure where coming this convintion there is something unexpressed in the XSX perfomance. Both hardware has their own costumization so more they learn to push better the XSX, same will happen for the ps5. Highly probable it will happen the contrary with the perfomance stay even more close because both hardware will be pushed to their limits to can't overtake by much the other counterparts.
 
Last edited:

winjer

Gold Member
And lower clocks with less aggressive turbo will also make a lot of chips more passable for Microsoft. I believe Series X yields probably benefit from the strategy, as Sony will get some semi-good processors, just not good enough for the load they will put them through. Liquid metal probably helps a lot though.

This said Sony is saving precious yield space with smaller dies so in the end it might just work out when it comes to cost.

Considering that RDNA2 has the capability of reaching clocks much higher than 2.2 Ghz, it's very unlikely that clocks speed will limit yields for the PS5.
 
Top Bottom