Xbox Scarlett APU Die Shot Analysis: Die Size Estimated at 401 mm2

  • 181 results
  • 1
  • 2
  • 3
  • 4
Avatar image for Pedro
Pedro

36585

Forum Posts

0

Wiki Points

0

Followers

Reviews: 64

User Lists: 0

#151 Pedro
Member since 2002 • 36585 Posts

@EG101 said:

I don't think your analogy works here.

Game graphics scale linearly with how much HP that engine (CPU & GPU) has while horse power on a car have a lot of other variables that affect performance.

For example things like, torque, gear ratios, traction, suspension, etc, etc all affect the performance of your vehicle.

Whether he nailed the analogy or not, the difference between the two systems is smaller than before.

Avatar image for Pedro
Pedro

36585

Forum Posts

0

Wiki Points

0

Followers

Reviews: 64

User Lists: 0

#152  Edited By Pedro
Member since 2002 • 36585 Posts
@BlackShirt20 said:

@Pedro: LOL ok. Hold on to that.

There is nothing to hold on to. It just is.

Avatar image for EG101
EG101

2024

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#153 EG101
Member since 2007 • 2024 Posts

@Pedro said:
@EG101 said:

I don't think your analogy works here.

Game graphics scale linearly with how much HP that engine (CPU & GPU) has while horse power on a car have a lot of other variables that affect performance.

For example things like, torque, gear ratios, traction, suspension, etc, etc all affect the performance of your vehicle.

Whether he nailed the analogy or not, the difference between the two systems is smaller than before.

2.8 TF's is the largest difference between consoles in the same gen.

This is mathematical fact. Not really open for debate.

Avatar image for xantufrog
xantufrog

12126

Forum Posts

0

Wiki Points

0

Followers

Reviews: 2

User Lists: 5

#154 xantufrog  Moderator
Member since 2013 • 12126 Posts

@EG101: and it's a small % delta, which is definitely relevant - how can you think it's not?

Avatar image for 04dcarraher
04dcarraher

23299

Forum Posts

0

Wiki Points

0

Followers

Reviews: 2

User Lists: 0

#155  Edited By 04dcarraher
Member since 2004 • 23299 Posts

@EG101 said:
@Pedro said:
@EG101 said:

I don't think your analogy works here.

Game graphics scale linearly with how much HP that engine (CPU & GPU) has while horse power on a car have a lot of other variables that affect performance.

For example things like, torque, gear ratios, traction, suspension, etc, etc all affect the performance of your vehicle.

Whether he nailed the analogy or not, the difference between the two systems is smaller than before.

2.8 TF's is the largest difference between consoles in the same gen.

This is mathematical fact. Not really open for debate.

Um no its not....

Xbox One was 1.3 TFLOP while PS4 was 1.84, and 1.3 TFLOP of 1.84 is a 30% difference..... And that's only 540 GFLOPS

X1X is 6 TFLOP while PS4 Pro is 4.2 TFLOPS. 4.2 TFLOPS of 6 is again 30% difference... and that's 1.8 TFLOPS

Now if we are to take the rumored XSX 12 TFLOPS and PS5's 9.2 is only a 24% difference while that's 2.8 TFLOPS

So the performance difference this time around will be less than the last two console versions percentage wise ..... Even though the TFLOP difference may be higher than the last two.

Avatar image for EG101
EG101

2024

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#156 EG101
Member since 2007 • 2024 Posts

Another thing is that you can do a lot more work with those 2.8 TF's than you can with the 1.8 TF's difference there was between the XB1X and PS4 Pro.

You guys saw the difference in performance there. Many games had 2 times the performance if Devs chose to take advantage of the XB1X HW.

Imagine what a whole 2.8 TF's will be able to do when you consider things like VRS and improved performance per TF of RDNA's Architecture.

2.8 TF's is a lot of useable performance.

Avatar image for xantufrog
xantufrog

12126

Forum Posts

0

Wiki Points

0

Followers

Reviews: 2

User Lists: 5

#157  Edited By xantufrog  Moderator
Member since 2013 • 12126 Posts

@04dcarraher: several of us have explained exactly that to him already. Not sure why he's not buying it - and of course my engine analogy would best if one assumes all other mechanical details are held constant but that's frankly overthinking it (of course this was probably done by him on purpose in order to dismiss my point outright because I didn't specify such a controlled experiment - my point only gets stronger in a situation where all aspects of the vehicle are 100% identical except HP)

Avatar image for EG101
EG101

2024

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#158  Edited By EG101
Member since 2007 • 2024 Posts

@xantufrog said:

@EG101: and it's a small % delta, which is definitely relevant - how can you think it's not?

I get the whole percentage delta but that's not how it works.

Let me ask you, which is greater 30% of 1 million or 10% of 1 billion?

If you say that 30% of a million is greater then you are correct in your assessment.

If 10% of 1 billion is greater then I'm correct.

Avatar image for xantufrog
xantufrog

12126

Forum Posts

0

Wiki Points

0

Followers

Reviews: 2

User Lists: 5

#159 xantufrog  Moderator
Member since 2013 • 12126 Posts

@EG101 said:
@xantufrog said:

@EG101: and it's a small % delta, which is definitely relevant - how can you think it's not?

I get the whole percentage delta but that's not how it works.

Let me ask you, which is greater 30% of 1 million or 10% of 1 billion?

If you say that 30% of a million is greater then you are correct in your assessment.

If 10% of 1 billion is greater then I'm correct.

That IS how it works - that's what everyone is trying to tell you.

Nobody is arguing with you that 2.8 is not a bigger number - we can all count, thanks. What we are trying to get you to understand is that as a performance difference the RATIO matters. It scales - the whole reason TFLOPS keep going up in the first place is because hardware is adapting to new software demands - given the exact same game on the exact same architecture (apples to apples) you will see a bigger performance difference between two GPUs with 2:1 TFLOPS ratio than two GPUs with a 1.5:1 TFLOPS ratio

I'm not going to keep explaining it - if you think I have lobsters for brains that's ok.

Avatar image for EG101
EG101

2024

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#160 EG101
Member since 2007 • 2024 Posts

@xantufrog said:
@EG101 said:
@xantufrog said:

@EG101: and it's a small % delta, which is definitely relevant - how can you think it's not?

I get the whole percentage delta but that's not how it works.

Let me ask you, which is greater 30% of 1 million or 10% of 1 billion?

If you say that 30% of a million is greater then you are correct in your assessment.

If 10% of 1 billion is greater then I'm correct.

That IS how it works - that's what everyone is trying to tell you.

Nobody is arguing with you that 2.8 is not a bigger number - we can all count, thanks. What we are trying to get you to understand is that as a performance difference the RATIO matters. It scales - the whole reason TFLOPS keep going up in the first place is because hardware is adapting to new software demands - given the exact same game on the exact same architecture (apples to apples) you will see a bigger performance difference between two GPUs with 2:1 TFLOPS ratio than two GPUs with a 1.5:1 TFLOPS ratio

I'm not going to keep explaining it - if you think I have lobsters for brains that's ok.

We'll see 2 to 3 years into next gen if the differences between a 400 GF console are indeed greater than the differences between 2.8 TF's.

We'll just disagree on this one for now.

Avatar image for EG101
EG101

2024

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#161 EG101
Member since 2007 • 2024 Posts
@04dcarraher said:
@EG101 said:
@Pedro said:
@EG101 said:

I don't think your analogy works here.

Game graphics scale linearly with how much HP that engine (CPU & GPU) has while horse power on a car have a lot of other variables that affect performance.

For example things like, torque, gear ratios, traction, suspension, etc, etc all affect the performance of your vehicle.

Whether he nailed the analogy or not, the difference between the two systems is smaller than before.

2.8 TF's is the largest difference between consoles in the same gen.

This is mathematical fact. Not really open for debate.

Um no its not....

Xbox One was 1.3 TFLOP while PS4 was 1.84, and 1.3 TFLOP of 1.84 is a 30% difference..... And that's only 540 GFLOPS

X1X is 6 TFLOP while PS4 Pro is 4.2 TFLOPS. 4.2 TFLOPS of 6 is again 30% difference... and that's 1.8 TFLOPS

Now if we are to take the rumored XSX 12 TFLOPS and PS5's 9.2 is only a 24% difference while that's 2.8 TFLOPS

So the performance difference this time around will be less than the last two console versions percentage wise ..... Even though the TFLOP difference may be higher than the last two.

Right the TF value is higher and therefore the differences will be greater.

Percentage means nothing in this comparison because 2.8 TF's is a huge difference.

The difference between XB1 1.4 TF's and PS4 1.8 TF's is 400 GF's.

Guess which one will make a bigger difference in games?

Avatar image for 04dcarraher
04dcarraher

23299

Forum Posts

0

Wiki Points

0

Followers

Reviews: 2

User Lists: 0

#162  Edited By 04dcarraher
Member since 2004 • 23299 Posts

@EG101 said:
@04dcarraher said:
@EG101 said:
@Pedro said:

Whether he nailed the analogy or not, the difference between the two systems is smaller than before.

2.8 TF's is the largest difference between consoles in the same gen.

This is mathematical fact. Not really open for debate.

Right the TF value is higher and therefore the differences will be greater.

Percentage means nothing in this comparison because 2.8 TF's is a huge difference.

The difference between XB1 1.4 TF's and PS4 1.8 TF's is 400 GF's.

Guess which one will make a bigger difference in games?

The differences between the two new gpu's will be smaller because the difference is only 24% vs 30% with the other revisions.....

Wrong.... Percentages tell you what to expect at best when it comes to the performance gap between the two gpus not taking into account other aspects ( ie bandwidth,memory etc).

The original X1 has a 1.3 TFLOP gpu not 1.4 TFLOP the 1.4 is the X1 S model.

Your splitting hairs...., we all know the XSX's gpu is going to beat the PS5's gpu if the rumors are true.... However the difference is going to smaller than what the X1 was to the PS4 or PS4 pro was to the X1X.

So again the 2.8 TFLOP gap aka 24% is a overall, is a smaller performance gap than what the other current consoles had between themselves.

Avatar image for 04dcarraher
04dcarraher

23299

Forum Posts

0

Wiki Points

0

Followers

Reviews: 2

User Lists: 0

#163  Edited By 04dcarraher
Member since 2004 • 23299 Posts

Ok....

Lets break it down a bit further for you to understand..... Lets look at the RX 5700 XT NITRO+ Special Edition which is overclocked to the point of being at 10.4TFLOPS. Now a stock RX 5700 with 7.9 TFLOPS. Those two gpu's have a 2.5 TFLOP difference in performance. Math dictates a 24% performance difference.

So that 2.5 TFLOP lead should be HUGE!!!!!!! buts its not....

Now with a 22 game test the overall performance gap of the gpu's is only 16% at 4k and 1440p max settings. Using 4k resolution BF5 has a 10 fps difference 75 vs 65 fps only 14% gap. Look at Rage 2 which 34 fps vs 32 FPS at 4k 6% difference. Or AC: Odyssey 30 fps vs 27 FPS 10% difference. Now lets look at a game where we see almost a 19% difference in framerate Devil May Cry 5, 58 vs 48 FPS.

If both consoles are using same memory type and amount with same bandwidth using the same RDNA architecture we are only going to see slight differences in graphics and performance because even with a 2.5 TFLOP advantage of the 5700XT nitro over the plain 5700 is 14% in framerates. So at most we will only see minor graphical differences ie select few high vs ultra setting or slightly less post process effects to gain back that small % of performance to reach that 30 or 60 FPS target.

Avatar image for Pedro
Pedro

36585

Forum Posts

0

Wiki Points

0

Followers

Reviews: 64

User Lists: 0

#164  Edited By Pedro
Member since 2002 • 36585 Posts

@EG101 said:

2.8 TF's is the largest difference between consoles in the same gen.

This is mathematical fact. Not really open for debate.

And mathematically it is the smallest difference between the two competing consoles.

Based on your other responses that followed, its clear that you lack the ability to understand what others have tried so hard for you to comprehend.

Avatar image for EG101
EG101

2024

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#165  Edited By EG101
Member since 2007 • 2024 Posts

@Pedro said:
@EG101 said:

2.8 TF's is the largest difference between consoles in the same gen.

This is mathematical fact. Not really open for debate.

And mathematically it is the smallest difference between the two competing consoles.

Based on your other responses that followed, its clear that you lack the ability to understand what others have tried so hard for you to comprehend.

I understand your point perfectly I simply disagree.

Also in the future do me a favor stay on topic and attack the argument.

Avatar image for EG101
EG101

2024

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#166 EG101
Member since 2007 • 2024 Posts
@04dcarraher said:

Ok....

Lets break it down a bit further for you to understand..... Lets look at the RX 5700 XT NITRO+ Special Edition which is overclocked to the point of being at 10.4TFLOPS. Now a stock RX 5700 with 7.9 TFLOPS. Those two gpu's have a 2.5 TFLOP difference in performance. Math dictates a 24% performance difference.

So that 2.5 TFLOP lead should be HUGE!!!!!!! buts its not....

Now with a 22 game test the overall performance gap of the gpu's is only 16% at 4k and 1440p max settings. Using 4k resolution BF5 has a 10 fps difference 75 vs 65 fps only 14% gap. Look at Rage 2 which 34 fps vs 32 FPS at 4k 6% difference. Or AC: Odyssey 30 fps vs 27 FPS 10% difference. Now lets look at a game where we see almost a 19% difference in framerate Devil May Cry 5, 58 vs 48 FPS.

If both consoles are using same memory type and amount with same bandwidth using the same RDNA architecture we are only going to see slight differences in graphics and performance because even with a 2.5 TFLOP advantage of the 5700XT nitro over the plain 5700 is 14% in framerates. So at most we will only see minor graphical differences ie select few high vs ultra setting or slightly less post process effects to gain back that small % of performance to reach that 30 or 60 FPS target.

That's a simple over clock not a SOC designed around the target specs.

Avatar image for Pedro
Pedro

36585

Forum Posts

0

Wiki Points

0

Followers

Reviews: 64

User Lists: 0

#168 Pedro
Member since 2002 • 36585 Posts

@EG101 said:
@Pedro said:

And mathematically it is the smallest difference between the two competing consoles.

Based on your other responses that followed, its clear that you lack the ability to understand what others have tried so hard for you to comprehend.

I understand your point perfectly I simply disagree.

Also in the future do me a favor stay on topic and attack the argument.

I am sorry but according to you "This is mathematical fact. Not really open for debate."

Avatar image for xantufrog
xantufrog

12126

Forum Posts

0

Wiki Points

0

Followers

Reviews: 2

User Lists: 5

#170  Edited By xantufrog  Moderator
Member since 2013 • 12126 Posts

@EG101: fair enough

Avatar image for 04dcarraher
04dcarraher

23299

Forum Posts

0

Wiki Points

0

Followers

Reviews: 2

User Lists: 0

#171 04dcarraher
Member since 2004 • 23299 Posts

does not matter..... the TFLOP numbers using same architecture allows us to gauge the processing power of the gpus between the sku's. It does not matter if one console has 56 CU clocked a 1.7 ghz and the other having 40 CU's at 2ghz. Or comparing a RX 5700XT OC'ed with 40 CU's vs a RX 5700 with 36 CU's at stock .... TFLOP numbers tell us where they land on the totem pole of performance within the generation.

If these consoles are using the same base architecture the numbers tell us what to expect. Like I stated earlier, this speculation is all based on them having equal footing with memory and bandwidth using the same base architecture aka RDNA 2.0.

Now that s not say that XSX could be using RDNA v1 with RDNA v2 features or if PS5 is straight RDNA 2.0 or vice versa.... Then we have no idea if the RT in both or either one are using a hybrid RT (which is still hardware based) or if it all dedicated hardware or if it's all AMD based or an 3rd party chip handling RT.

But when it comes the TFLOP numbers the percentage difference is smaller this time around than the last two releases even though there is a 2.8 TFLOP difference.....

Avatar image for brimmul777
brimmul777

4266

Forum Posts

0

Wiki Points

0

Followers

Reviews: 42

User Lists: 2

#172 brimmul777
Member since 2011 • 4266 Posts

I like shiny things. 🙂

Avatar image for Martin_G_N
Martin_G_N

1932

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#173  Edited By Martin_G_N
Member since 2006 • 1932 Posts

The power difference is still not big enough to save MS. Sony's first party titles will still be the graphics kings on consoles, and third party titles will have similar performance on both consoles, with minor differences in graphics settings. A lot of the early games will be cross gen, which means they won't properly take advantage of the new hardware anyways.

Most games later on will target 4K 30FPS with beefy new graphics and physics tech. Some games will target 60FPS, and if the PS5 struggles with some minor dips, then devs can opt for the checkerboard rendering which when done right will look almost identical to the native image. At least for regular consumers. I think the performance chosen for the PS5 is just right, to keep the price at the right level, and at the same time give developers the option to create bigger and better games. Most people will still play games at 1080p.

There is nothing the Series X can do that the PS5 cannot, however, the PS4, pro, X1, and X1X will struggle to run certain next gen games without developers rebuilding the game and adding load screens. Because of the fast SSD, and faster CPU.

Avatar image for EG101
EG101

2024

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#174  Edited By EG101
Member since 2007 • 2024 Posts

@Pedro said:
@EG101 said:
@Pedro said:

And mathematically it is the smallest difference between the two competing consoles.

Based on your other responses that followed, its clear that you lack the ability to understand what others have tried so hard for you to comprehend.

I understand your point perfectly I simply disagree.

Also in the future do me a favor stay on topic and attack the argument.

I am sorry but according to you "This is mathematical fact. Not really open for debate."

Percentages mean nothing without context.

If the rumored specs are true there is a difference of 2.8 TF's which is the largest amount of Flops between same gen consoles.

A difference of 2.8 TF's is factually larger than any other gap between direct competitors.

What people here are trying to convince me is that the gap will be less noticeable than the gaps between past consoles.

This is a legit argument that I disagree with. I believe that a lot of work can be done with those Flops as long as the HW was designed around maximizing that spec.

Avatar image for ronvalencia
ronvalencia

28449

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#175  Edited By ronvalencia
Member since 2008 • 28449 Posts

@Pedro said:

@ronvalencia: You spew all of that garble and nothing changed. The difference between the two consoles will still be negligible.

Your arguments doesn't add to this topic.

Facts

R9-390X has effectively two Pitcairn XT glued together with about twice the performance.

R9-390X Hawaii XT (44 CU, quad geometry-raster units, 64 ROPS, 512-bit bus) has about twice the hardware over R9-270X (20 CU, dual-geometry-raster units, 32 ROPS, 256-bit bus).

-----------

If AMD created Three Shader engines GCN, 72 percent level from R9-390X's 99 percent would be half way point between R9-270X and R9-390X.

Avatar image for tormentos
tormentos

29532

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#176 tormentos
Member since 2003 • 29532 Posts
@EG101 said:

Percentages mean nothing without context.

If the rumored specs are true there is a difference of 2.8 TF's which is the largest amount of Flops between same gen consoles.

A difference of 2.8 TF's is factually larger than any other gap between direct competitors.

What people here are trying to convince me is that the gap will be less noticeable than the gaps between past consoles.

This is a legit argument that I disagree with. I believe that a lot of work can be done with those Flops as long as the HW was designed around maximizing that spec.

Say what?😂😂

Dude STOP.

Here look at the difference between the RX5700XT and the RX5700, 1.8 TF difference between both and that only amount to 7 freaking frame per second 7 frames how can 1.8TF just amount to 7 FPS more? You do know the PS4 had just over 500Gflops on the xbox one,yet commanded leads in frames in several games bigger than 7FPS?

Loading Video...
Loading Video...

Explain to me how 500+Gflops more on PS4 is enough to generate gaps as big as this,but 1.8TF can only produce 7FPS gap?

I chose Forza because it is a console oriented game.

To what amount 1.8TF more on Navi 5700Xt vs 5700 7FPS on Forza.

The RX5700 has 22% higher flop count,the PS4 has 40% higher flops,yet in actual flop the the PS4 gap over the xbox one fit 3+ times inside 1.8TF.

So what produce a bigger gap 40% more flops,or 1.8TF which represent 3+ times more than the actual 40% gap on PS4?

You try to pass this crap as well as other lemming with the xbox one X,in that time the argument was that the gap between the xbox one X and pro was a complete PS4,in a pathetic attempt to make the gap seen bigger than it was.

So yeah % does matter more than actual flops,so yeah the gap is the smallest between the ps4 xbox one and PS4 Pro xbox one X scenario.

Avatar image for ronvalencia
ronvalencia

28449

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#177  Edited By ronvalencia
Member since 2008 • 28449 Posts

@tormentos said:
@ronvalencia said:

XSX GPU can be 60 CU with the following configuration

Config 1

Three RDNA Shader Engines. Each RDNA Shader Engine has 20 CU. Two CUs are closely glued together as DCU (dual CU).

Config 2

Two RDNA Shader Engines. Each RDNA Shader Engine has 30 CU. Two CUs are closely glued together as DCU (dual CU).

Backgound information on RDNA GPU design framework

Two CUs are binded together as DCU (Dual CU).

Config 1 has the benefit with scaling RBE (Render Balckend) and Geometry Prim Units scaling with CU count.

Config 2 has diminishing returns with additional CUs. Fury X/Vega raster graphics IPC degradation (which is not a problem for server TFLOPS workloads).

Known RDNA deployment

NAVI 10's Shader Engine has 20 CU. NAVI 10 has two Shader Engines

NAVI 14's single Shader Engine has 24 CU.

RX-5600 XT has NAVI 10 with 192 bits bus which is first time AMD was able to follow NVIDIA's Pascal/Turing 192 bit bus design, hence extra BOM cost flexibility for AMD.

https://hothardware.com/news/microsoft-xbox-series-x-holiday-2020-zen-2-amd-rdna

Getting down to brass tacks, Microsoft says that it is using a "custom designed" AMD Zen 2-based processor (although early reports have indicated that it will be based on an 8-core/16-thread Ryzen 3000 design) and "next generation" RDNA graphics architecture

Microsoft claims "next generation" RDNA graphics architecture which points to "RDNA 2".

RDNA 2 includes 7nm+ EUV with 20 percent increase density.

https://www.techpowerup.com/254656/amd-zen3-to-leverage-7nm-euv-for-20-transistor-density-increase?cp=2

7 nm+ EUV (extreme ultraviolet) silicon fabrication node at TSMC, which promises a significant 20 percent increase in transistor densities compared to the 7 nm DUV (deep ultraviolet) node

325 mm2 with "7nm+ EUV" roughly equalvent to 390 mm2 size GPU under 1st gen 7nm DUV node

Microsoft confirm their Scarlet APU has "RDNA 2".

And hereeee we goooo with your usual speculation did MS CLAIMED RDNA 2?

OPENLY?

From what i read on resetera from leaks is RDNA 1 with RDNA 2 features the same is say about the PS5.

And is not new we know the Pro was polaris with vega features is not new.

This leaks are about the development kit not about retail units,which mean if the dev kit has 56CU the final hardware will have less,just like the XBO X has 4 less than the devkit.

So it can't be 60CU.

1. MS claimed "next generation RDNA" after RX-5700 XT RDNA v1 was released, hence MS has implied "RDNA 2" claims.

From https://www.overclock3d.net/news/gpu_displays/amd_s_rdna_2_architecture_will_deliver_major_performance_efficiency_improvements_-_rumour/1

Comes from the same AMD 's RDNA road map press deck.

https://news.xbox.com/en-us/2019/12/12/microsoft-unveils-xbox-series-x/

Powered by our custom-designed processor leveraging the latest Zen 2 and next generation RDNA architecture from our partners at AMD, Xbox Series X will deliver hardware accelerated ray tracing and a new level of performance never before seen in a console. Additionally, our patented Variable Rate Shading (VRS) technology will allow developers to get even more out of the Xbox Series X GPU and our next-generation SSD will virtually eliminate load times and bring players into their gaming worlds faster than ever before.

Hardware accelerated ray tracing and Variable Rate Shading (VRS) features are also on NVIDIA's Turing RTX GPUs. These features doesn't exist on PC's RDNA v1

-----------------------

2. For XSX GPU retail context, MS claimed twice of X1X GPU "when doing the math".

RDNA and GCN math calculators are similar when raster hardware and memory bandwidth are not the bottleneck factors.

Scale Hawaii R9-390's 5.1 TFLOPS into RX-5700 XT's 9.75 TFLOPS, it would land between Vega 64 and GTX 1080 Ti i.e. 112 percent

Hawaii 40 CU raster IPC is similar or closer to NAVI 10 40 CU raster IPC.

X1X GPU's results are Hawaii Pro like raster IPC with overclocking (e.g. 1172 Mhz) and design improvements e.g. blending and alpha effects, Polaris DCC (mitigates missing 128 bit bus from Hawaii's 512 bit bus).

Both Hawaii and Polaris 10/20 has quad geometry improvements i.e. both GPUs has quad Shader Engines.

R9-390's 5.1 TFLOPS raster IPC beaten RX-470's 4.9 TFLOPS Polaris raster IPC. Don't underestimate TFLOPS scaling with memory bandwidth scaling which both factors can reduce render times.

Avatar image for ronvalencia
ronvalencia

28449

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#178  Edited By ronvalencia
Member since 2008 • 28449 Posts

@tormentos said:
@EG101 said:

Percentages mean nothing without context.

If the rumored specs are true there is a difference of 2.8 TF's which is the largest amount of Flops between same gen consoles.

A difference of 2.8 TF's is factually larger than any other gap between direct competitors.

What people here are trying to convince me is that the gap will be less noticeable than the gaps between past consoles.

This is a legit argument that I disagree with. I believe that a lot of work can be done with those Flops as long as the HW was designed around maximizing that spec.

Say what?😂😂

Dude STOP.

Here look at the difference between the RX5700XT and the RX5700, 1.8 TF difference between both and that only amount to 7 freaking frame per second 7 frames how can 1.8TF just amount to 7 FPS more? You do know the PS4 had just over 500Gflops on the xbox one,yet commanded leads in frames in several games bigger than 7FPS?

Explain to me how 500+Gflops more on PS4 is enough to generate gaps as big as this,but 1.8TF can only produce 7FPS gap?

I chose Forza because it is a console oriented game.

To what amount 1.8TF more on Navi 5700Xt vs 5700 7FPS on Forza.

The RX5700 has 22% higher flop count,the PS4 has 40% higher flops,yet in actual flop the the PS4 gap over the xbox one fit 3+ times inside 1.8TF.

So what produce a bigger gap 40% more flops,or 1.8TF which represent 3+ times more than the actual 40% gap on PS4?

You try to pass this crap as well as other lemming with the xbox one X,in that time the argument was that the gap between the xbox one X and pro was a complete PS4,in a pathetic attempt to make the gap seen bigger than it was.

So yeah % does matter more than actual flops,so yeah the gap is the smallest between the ps4 xbox one and PS4 Pro xbox one X scenario.

Your argument with 5700 vs 5700 XT is valid when PS5 has XSX's at least 320 bit GDDR-14000 memory bus.

Are you claiming PS5 has 320 bit GDDR-14000 memory bus?

XBO has memory architecture complexity with non-transparent (manual programming effort) frame buffer load balancing prior to middle-ware update.

Both X1X and PS4 Pro has simple single unified memory architecture with X1X including optional hyper fast 2MB render cache equipped ROPS micro-tile render advantage. Both PS4 Pro and X1X GPUs has 2MB L2 cache which is connected to geometry and TMUs.

XBO vs PS4 based arguments are useless when PS5 and XSX has simple single unified memory architecture and NAVI 10's atleast 4MB L2 cache equipped ROPS micro-tile render option.

Let's see MS stupid enough to scale CU (TFLOPS) without scaling memory bandwidth, L2 cache and raster hardware I/O. MS has increased memory bandwidth and cache storage size with X1X GPU's 6 TFLOPS.

PS4 Pro was the stupid design when it scaled it's TFLOPS by 2.3X and raw memory bandwidth by 1.23X (not factoring Polaris DCC) from PS4.

Avatar image for tormentos
tormentos

29532

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#179 tormentos
Member since 2003 • 29532 Posts

@ronvalencia said:

Your argument with 5700 vs 5700 XT is valid when PS5 has XSX's at least 320 bit GDDR-14000 memory bus.

Are you claiming PS5 has 320 bit GDDR-14000 memory bus?

XBO has memory architecture complexity with non-transparent (manual programming effort) frame buffer load balancing prior to middle-ware update.

Both X1X and PS4 Pro has simple single unified memory architecture with X1X including optional hyper fast 2MB render cache equipped ROPS micro-tile render advantage.

XBO vs PS4 based arguments are useless when PS5 and XSX has simple single unified memory architecture and NAVI 10's atleast 4MB L2 cache equipped ROPS micro-tile render option.

NO at all.

Because even with more bandwidth the RX5700 still has less power.

It is unkown the exact bandwidth of the PS5,rumors claim 520/gbs but is not confirmed.

It doesn't matter the xbox one X has 1.8TF on PS4 pro which more or less is the same gap between the xbox one and PS4 and in both cases both have show gaps bigger than 100%,so regardles of been 500+ Gflops on the PS4 case or 1.8TF on the xbox one X case the gap is more or less the same,even that 1.8TF is more than 3 times bigger than 500+Gflops.

So yeah even if the xbox series X end been stronger the gap been show by % is the smallest.

Avatar image for ronvalencia
ronvalencia

28449

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#180  Edited By ronvalencia
Member since 2008 • 28449 Posts

@tormentos said:
@ronvalencia said:

Your argument with 5700 vs 5700 XT is valid when PS5 has XSX's at least 320 bit GDDR-14000 memory bus.

Are you claiming PS5 has 320 bit GDDR-14000 memory bus?

XBO has memory architecture complexity with non-transparent (manual programming effort) frame buffer load balancing prior to middle-ware update.

Both X1X and PS4 Pro has simple single unified memory architecture with X1X including optional hyper fast 2MB render cache equipped ROPS micro-tile render advantage.

XBO vs PS4 based arguments are useless when PS5 and XSX has simple single unified memory architecture and NAVI 10's atleast 4MB L2 cache equipped ROPS micro-tile render option.

NO at all.

Because even with more bandwidth the RX5700 still has less power.

It is unkown the exact bandwidth of the PS5,rumors claim 520/gbs but is not confirmed.

It doesn't matter the xbox one X has 1.8TF on PS4 pro which more or less is the same gap between the xbox one and PS4 and in both cases both have show gaps bigger than 100%,so regardles of been 500+ Gflops on the PS4 case or 1.8TF on the xbox one X case the gap is more or less the same,even that 1.8TF is more than 3 times bigger than 500+Gflops.

So yeah even if the xbox series X end been stronger the gap been show by % is the smallest.

Memory bandwidth is just a tube for data (think of water) to travel on. Larger tube is useless when water source is low.

Higher geometry I/O = lower render time.

Higher texture I/O = lower render time.

Higher TFLOPS = lower render time.

Higher raster I/O = lower render time.

Higher data transfer I/O= lower render time.

Each stage lowers render time and has dependencies with other stages e.g. higher memory bandwidth I/O needs higher classic GPU hardware I/O which feeds CUs, then dumps to raster hardware I/O, and then memory bandwidth I/O. Hint: graphics pipeline dependencies.

XBO vs PS4, XBO has ESRAM vs DDR3 interference. Useless for PS5 vs XSX discussion.

X1X vs PS4 Pro, both consoles have similar large flat memory address space architecture with different memory bandwidth and GPU IP selections.

-----------

XBO S vs GTX 750 Ti with recent games

Loading Video...

Video's debate points

  • The slowest section can gimp performance
  • Balance design.

XBO S has better balance when compared to GTX 750 Ti

PS4 has better balance when compared to XBO S

X1X has better balance when compared to PS4 Pro

Avatar image for tormentos
tormentos

29532

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#181 tormentos
Member since 2003 • 29532 Posts

@ronvalencia:

Hardware accelerated ray tracing and Variable Rate Shading (VRS) features are also on NVIDIA's Turing RTX GPUs. These features doesn't exist on PC's RDNA v1

This is the only thing i will reply to,i love when you play the blind chicken and start acting like it has to be RDNA2 mandatory.

Polaris doesn't have FP16 yet the PS4 Pro has it,is a VEGA feature,just like the PS4 had 8 aces when the 7870 doesn't have that either.

The fact that Ray Tracing and VRS are inside the series X doesn't mandatory mean they are RDNA2,those features can be inside RDNA1 by custom design just like sony did with Aces and FP16,by the way the PS5 also is say to have hardware acelerated ray tracing,and VRS as well.

So yeah both can be RDNA1 or 2 but having those features doesn't mandatory means they are RDNA2 and you know it.

Avatar image for Pedro
Pedro

36585

Forum Posts

0

Wiki Points

0

Followers

Reviews: 64

User Lists: 0

#182 Pedro
Member since 2002 • 36585 Posts

@ronvalencia: And the same rules applied like it did before. More garble and nothing of substance.

Avatar image for BlackShirt20
BlackShirt20

1343

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#183 BlackShirt20
Member since 2005 • 1343 Posts

@Pedro: keep that faith. Will only disappoint you even more.

Avatar image for BlackShirt20
BlackShirt20

1343

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#184 BlackShirt20
Member since 2005 • 1343 Posts

@tormentos: negative. XSX will statistically be a better console. That’s all I have said. It is all but confirmed the XSX is the powerful system. You can cry all you want, but all the leaks are from AMD files. Photos of the actual chip, architecture videos ect.

Avatar image for Pedro
Pedro

36585

Forum Posts

0

Wiki Points

0

Followers

Reviews: 64

User Lists: 0

#185 Pedro
Member since 2002 • 36585 Posts

@BlackShirt20 said:

@tormentos: negative. XSX will statistically be a better console. That’s all I have said. It is all but confirmed the XSX is the powerful system. You can cry all you want, but all the leaks are from AMD files. Photos of the actual chip, architecture videos ect.

Can you post the official links to the PS5 and Series X to validate your claim?

Avatar image for ronvalencia
ronvalencia

28449

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#186  Edited By ronvalencia
Member since 2008 • 28449 Posts

@tormentos said:

@ronvalencia:

Hardware accelerated ray tracing and Variable Rate Shading (VRS) features are also on NVIDIA's Turing RTX GPUs. These features doesn't exist on PC's RDNA v1

This is the only thing i will reply to,i love when you play the blind chicken and start acting like it has to be RDNA2 mandatory.

Polaris doesn't have FP16 yet the PS4 Pro has it,is a VEGA feature,just like the PS4 had 8 aces when the 7870 doesn't have that either.

The fact that Ray Tracing and VRS are inside the series X doesn't mandatory mean they are RDNA2,those features can be inside RDNA1 by custom design just like sony did with Aces and FP16,by the way the PS5 also is say to have hardware acelerated ray tracing,and VRS as well.

So yeah both can be RDNA1 or 2 but having those features doesn't mandatory means they are RDNA2 and you know it.

Read https://gpuopen.com/using-sub-dword-addressing-on-amd-gpus-with-rocm/ before you embarrass yourself.

Polaris has dual packed FP16 per FP32 slot which consumes an ALU, hence reducing two register usage for a single register.

CU has four lanes. Each lane has four SIMD4.

CU has 16 SIMD total or 64 FP32 operations per cycle i.e. 16 x SIMD4 = 64 data processing per cycle.

Programmer's wave64 has 16 SIMD payload.

Tonga/Fiji/Polaris's pack dual packed FP16 doesn't expand CU's operation rate, hence it stays at 64 operations per cycle per CU.

Vega also has dual packed FP16 per FP32 register slot, but it can expand to 128 operations per cycle a.k.a. double rate FP16 when the programmer fully populates wave64 with dual FP16s.

MS didn't select Vega's double rate FP16 (aka rapid pack math) feature while keeping Polaris pack FP16 feature. MS selected ROPS with multi-MB cache instead which is similar to Vega ROPS design.

Vega TMU and ROPS connected to 4MB L2 cache

X1X ROPS has 2MB cache while TMU has 2MB L2 cache i.e. missing unified Vega L2 cache link.

Sony selected Vega's double rate feature and keeps the old behavior GCN ROPS for PS4 Pro.

Both MS and Sony has been presented with Polaris baseline and Vega IP.

Full Vega update has rapid pack math and ROPS improvements.

Both MS and Sony have different slices from Vega IP selection.

MS specifically confirms Vega's double rate FP16 (aka rapid pack math) feature was NOT selected for X1X.

------------

X1X GPU's 150 to 160 watts at 6 TFLOPS is about half of Vega 64's 12.58 TFLOPS and 295 watts, hence MS has Vega's performance per watt enhancements. RX-480's performance per watt can't scale into Vega 64's 12.58 TFLOPS and 295 watts TDP.

XBO has DX12 command processor (micro-coding engine), dual graphics command processor units and two ACE units (16 queues).

Sony has 8 ACE units for SPE middleware port, hence Sony has different requirements.

Avatar image for ronvalencia
ronvalencia

28449

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#187 ronvalencia
Member since 2008 • 28449 Posts

@Pedro said:

@ronvalencia: And the same rules applied like it did before. More garble and nothing of substance.

You have been reported for derailing this thread.

Avatar image for SecretPolice
SecretPolice

35406

Forum Posts

0

Wiki Points

0

Followers

Reviews: 0

User Lists: 0

#188 SecretPolice
Member since 2007 • 35406 Posts

:P

Avatar image for mclarenmaster18
MclarenMaster18

2103

Forum Posts

0

Wiki Points

0

Followers

Reviews: 29

User Lists: 5

#189 MclarenMaster18
Member since 2014 • 2103 Posts

To me, Xbox Series X will be the best gaming console for 9th gen in my opinion.

Avatar image for ronvalencia
ronvalencia

28449

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#190  Edited By ronvalencia
Member since 2008 • 28449 Posts

@tormentos said:
@ronvalencia said:

Your argument with 5700 vs 5700 XT is valid when PS5 has XSX's at least 320 bit GDDR-14000 memory bus.

Are you claiming PS5 has 320 bit GDDR-14000 memory bus?

XBO has memory architecture complexity with non-transparent (manual programming effort) frame buffer load balancing prior to middle-ware update.

Both X1X and PS4 Pro has simple single unified memory architecture with X1X including optional hyper fast 2MB render cache equipped ROPS micro-tile render advantage.

XBO vs PS4 based arguments are useless when PS5 and XSX has simple single unified memory architecture and NAVI 10's atleast 4MB L2 cache equipped ROPS micro-tile render option.

NO at all.

Because even with more bandwidth the RX5700 still has less power.

It is unkown the exact bandwidth of the PS5,rumors claim 520/gbs but is not confirmed.

It doesn't matter the xbox one X has 1.8TF on PS4 pro which more or less is the same gap between the xbox one and PS4 and in both cases both have show gaps bigger than 100%,so regardles of been 500+ Gflops on the PS4 case or 1.8TF on the xbox one X case the gap is more or less the same,even that 1.8TF is more than 3 times bigger than 500+Gflops.

So yeah even if the xbox series X end been stronger the gap been show by % is the smallest.

PS5 memory bandwidth rumors have ranged from 448 GB/s to 512 GB/s ("according to some test").

320 bit with GDDR6-12000 has 480 GB/s

-----

288 bit with GDDR-14000 has 504 GB/s

320 bit with GDDR-14000 has 560 GB/s

352 bit with GDDR-14000 has 616 GB/s

384 bit with GDDR-14000 has 672 GB/s

------

256 bit with GDDR6-16000 has 512 GB/s (assume RDNA v2 RX-5700 XT replacement i.e. RX-6700 XT would have GDDR6-16000)

Avatar image for ronvalencia
ronvalencia

28449

Forum Posts

0

Wiki Points

0

Followers

Reviews: 1

User Lists: 0

#191 ronvalencia
Member since 2008 • 28449 Posts

http://www.redgamingtech.com/investigating-the-ps5-xbox-series-x-gpu-architectures/

New specs PS5 Paste Bin has 40 CU at 2000Mhz (10.25 TFLOPS) and 512 GB/s.

Known RDNA Shader Engine has reached 24 CU, hence dual RDNA Shader Engine can support 44 CU configuration i.e. 40 CU active with 4 CU disabled for improved yields.