They got this number either by using a mish-mash of metrics or just CPU power, depending on who you ask within MS itself.
So are we gonna take this quote literally and seriously?
A cut down zen 2 would be 40 mm2. thats a 291mm2 apu with maybe 20 mm2 reserved for i/o stuff. 311mm2 max. smaller than the pro, the base ps4, base xb1 and the xb1x. not bigger.If somebody things ps5 will have 36 cu(40cu with 4 disabled) because backward compatibility then he is wrong. They will go with this number because rx5700xt with 40cu is 251mm^2 at 7nm and its even little bigger than expected for console gpu and sony don't want to make the same mistake as with ps3 (sell consoles with big loses).
No, 4 times as powerful references CPU and not GPU. Oluasc also gave you some good indicators what to expect for each system component improvement.So are we gonna take this quote literally and seriously?
So are we gonna take this quote literally and seriously?
EG: If you let me back I'll be here. Back to Scarlett... you mentioned it has four times the power of Xbox One X, which certainly sounds good. But what does that mean?
Matt Booty: It's a few things - it's the combination of speed, not just of the SSD but of the processor, the performance of the GPU and RAM, but we're also in a world where speed is starting not to matter.
Microsoft is calling Scarlett "a bigger leap than any generation we've done before," specifying that from a "pure processing perspective, [Scarlett is] four times more powerful than Xbox One X." Today, Variety was able to confirm with Microsoft that this number was specifically in reference to Scarlett's CPU, and not it's graphical horsepower, which we still know very little about.
Well now I don't know who to believe
Actually, they will be pushing it as far as they can under their constraints (budget, time, etc..). Whatever gets released late 2020 will be the foundation for the rest of the gen, so it makes more sense to get that as high as possible.Don't expect Sony or MS to push the hardware as far as they can. They'll want to leave some buffer for the pro models in 3-4 years.
This makes no sense. Its like you don't understand why consoles like the Pro/X even exists.Don't expect Sony or MS to push the hardware as far as they can. They'll want to leave some buffer for the pro models in 3-4 years.
believe the CEO of Microsoft game studios Phil Spencer himself.
Then by that definition, do Forza Horizon 3 & 4 have ray tracing as well? Both games also use a form of voxel based GI solution. Furthermore, as you said, voxelization has a very different performance trade off and scene accuracy against ray-geometry intersection. This isn't being stuck in naming semantics when the methods to achieve the end results are very different. You might as well be calling SSAO and HBAO+ the same thing then. You're seriously the first person I've seen calling voxel based GI ray tracing.You're getting stuck in naming semantics.
It's a dynamically generated ray-search acceleration structure that you can use to generate GI, reflections, or just render the entire scene (completely skip rasterization). So yes, you can trace rays for every pixel of the scene on screen.
Voxelizing instead of generating a BVH tree has its own (and very different) performance/memory trade-offs, but ultimately serves the same purpose. It doesn't directly compute ray-triangle-intersections, but that's never been a pre-requisite for RT.
so yea, connection between Oberon to Ariel is now established.
so gonzalo was really the PS5 all along then.
so yea, connection between Oberon to Ariel is now established.
so gonzalo was really the PS5 all along then.
Lol searching my images folder for Colbert's graphic of CUs, clock speeds, flops.
So would it have to be 36 CU for BC or 40? 40 with 4 off or 44 with 4 off?
Either way I think with 16GB, far better CPUs, and a 9Tflop Navi GPU (and SSD!) next gen will be a revelation. I think people are sleeping on how big an upgrade it could be.
Sony are using a custom GPU, I don't see it being shared with other devices. Anything that uses Ariel is most definitely something from Sony.How does this prove PS5 is Gonzalo? There's no logical following here. This is the first time we've ever heard of Oberon also.
My question is:
If Oberon = Ariel, Gonzalo = Ariel, Flute = Ariel.... then how do we know these are all the same chip?
Has AMD ever previously used multiple codenames for the same GPU/APU?
As if not, if Ariel is the GPU codename, it's possible the three above are three different APU products with the same GPU basis.... which wouldn't point to a games console (not Sony's at least).
Oh sure. Sounds like a great machine. But I am curious if Sony's design might have been partly limited by the requirement for BC and their less abstracted approach meaning they don't have the flexibility MS do in that area so need a little hardware assistance
It feels like a 2GHz (higher than any shipping AMD Navi right now) is way high for a console and especially one that the manufacturer hasn't always been they greatest around cooling. But did BC force their hand a little? Would they have preferred more CUs?
Or is this all roughly where they were going to land anyway
If 40cu 1.8ghz is doable on 7nm (simulating gonzalo thread) probably 36cu 2ghz alsoA cut down zen 2 would be 40 mm2. thats a 291mm2 apu with maybe 20 mm2 reserved for i/o stuff. 311mm2 max. smaller than the pro, the base ps4, base xb1 and the xb1x. not bigger.
there is also a pretty good chance that 2.0 ghz pretty much guarantees its on 7nm+ as AMD wasnt comfortable clocking their 7nm desktop GPUs at 2.0 ghz, even the anniversary edition which topped out at 1.98ghz.
and if its on 7nm+, 315mm2 will be the same as 350-360mm2 7nm apu which means it will definitely be more than 40 cus.
Oberon is the system codename, gonzalo is the APU, Flute the CPU and Ariel the GPU.How does this prove PS5 is Gonzalo? There's no logical following here. This is the first time we've ever heard of Oberon also.
My question is:
If Oberon = Ariel, Gonzalo = Ariel, Flute = Ariel.... then how do we know these are all the same chip?
Has AMD ever previously used multiple codenames for the same GPU/APU?
As if not, if Ariel is the GPU codename, it's possible the three above are three different APU products with the same GPU basis.... which wouldn't point to a games console (not Sony's at least).
I don't understand at all the relation of the number of CU and BC?
A cut down zen 2 would be 40 mm2. thats a 291mm2 apu with maybe 20 mm2 reserved for i/o stuff. 311mm2 max. smaller than the pro, the base ps4, base xb1 and the xb1x. not bigger.
there is also a pretty good chance that 2.0 ghz pretty much guarantees its on 7nm+ as AMD wasnt comfortable clocking their 7nm desktop GPUs at 2.0 ghz, even the anniversary edition which topped out at 1.98ghz.
and if its on 7nm+, 315mm2 will be the same as 350-360mm2 7nm apu which means it will definitely be more than 40 cus.
No, flute is the platform.Oberon is the system codename, gonzalo is the APU, Flute the CPU and Ariel the GPU.
If we know that Oberon is PS5 so it's an almost finished product by now with all the right spec.
Flute is the platform name. Komachi himself said that platform code names at AMD are named after instruments. I would consider your interpretation as incorrect.Oberon is the system codename, gonzalo is the APU, Flute the CPU and Ariel the GPU.
If we know that Oberon is PS5 so it's an almost finished product by now with all the right spec.
I like that theory very muchNo, flute is the platform.
What I think, is Gonzalo is devkit v1 with 40CU@1800MHz and Oberon is new devkit with new cooling with 36CU@2000MHz
Marketing wise it's extremely smart to have a rumour or leak like this considering Sony hasn't barely talked about ps5 at all this year. It moves focus away from Ms and let's everybody talk about ps5. Sony isn't spending a penny on marketing and the hype is already way up.
I think all of this is intentional.
No, the source of these leaks is AMD fucking up and putting private data for many different chips, not just Ariel and related in a place where komachi managed to access.Marketing wise it's extremely smart to have a rumour or leak like this considering Sony hasn't barely talked about ps5 at all this year. It moves focus away from Ms and let's everybody talk about ps5. Sony isn't spending a penny on marketing and the hype is already way up.
I think all of this is intentional.
in general the whole fast but narrrow vs slow and wide:
for fast and narrow solution:
cheaper to produce the APU and smaller in size
in cases of close TF number, the fast but narrow will often beat the slow and wide solution.
higher pixel throughput.
more performance that can be used for the BC boost mode.
slow and wide:
lower TDP, could use that power consumption on other parts of the system, like a higher clocked CPU, or faster memory.
using cheaper cooling solution probably.
No, the source of these leaks is AMD fucking up and putting private data for many different chips, not just Ariel and related in a place where komachi managed to access.
He actually held back a lot of the details to not risk themselves, but this isn't a Sony only leak, Arden is listed there for example.
An intentional leak will be more along the lines of the thurrott articles that we had in late 2018/early 2019.
No, they said that it was public in a tweet, but didnt link because they didnt want to risk themselves.It's more likely that Komachi has a source inside AMD. Which would mean AMD didn't screw up, rather someone is leaking inside info from within AMD.
I'd argue it's even better, since a PS4 with 2.4TF GPU would be around the GTX 660 Ti in performance. This time around, assuming the 9.2TF rumor is true, we have the equivalent of a 5700XT or RTX 2070 on our hands, the latter of which would be a match for the GTX 670 back in 2013.My thought, ps5 with 9.2tf navi in 2020 it would be like ps4 with 2.4tf gcn in 2013, quite impressive + now we have much faster cpu and ssd.
It's probably more than X2 GPU power actually. Even the 7.5TF RX 5700 (based on gaming clock) is more powerful than Vega64 which is 12.5 GCN TF. A 2Ghz 36CU GPU is more or less comparable to a 15TF GCN GPU which is 2.5X the power of the X GPU. If we get more than 36CU, I wouldn't even imagined that kind of power in my wildest dreams.Probably CPU alone or overall.
Expect things like this:
CPU: 4X over X
GPU: 2X over X
RAM: 1.5X over X
SSD: 40X over X
Depands how you look, we simplify here that 2x more tflops = 2x more performance but it's rarely true (bigger navi has worse perf/tf ratio).It's probably more than X2 GPU power actually. Even the 7.5TF RX 5700 (based on gaming clock) is more powerful than Vega64 which is 12.5 GCN TF. A 2Ghz 36CU GPU is more or less comparable to a 15TF GCN GPU which is 2.5X the power of the X GPU. If we get more than 36CU, I wouldn't even imagined that kind of power in my wildest dreams.
No, that estimation is wrong.It's probably more than X2 GPU power actually. Even the 7.5TF RX 5700 (based on gaming clock) is more powerful than Vega64 which is 12.5 GCN TF. A 2Ghz 36CU GPU is more or less comparable to a 15TF GCN GPU which is 2.5X the power of the X GPU. If we get more than 36CU, I wouldn't even imagined that kind of power in my wildest dreams.
No, flute is the platform.
What I think, is Gonzalo is devkit v1 with 40CU@1800MHz and Oberon is new devkit with new cooling with 36CU@2000MHz
Thx i missed the Flute tweet.Flute is the platform name. Komachi himself said that platform code names at AMD are named after instruments. I would consider your interpretation as incorrect.
You are also wrong here, because big navi is so inefficient, 9.3tf(on avarage) 5700xt is 1.14x faster than 12.58tf vega64 so it's like 14.34tf vega64, 9.2tf navi would be like 14tf vega 64.(...)9.2TF on Navi would perform like 12.8TF on Polaris, so basically 3x faster than PS4 Pro's GPU and 2.13x faster than X1X GPU.
Can you like to those tests?No, that estimation is wrong.
The Vega 64 TF number is not representative of its performance because it has a major bottleneck due to 64CU usage.
We have IPC tests done to the architecture (by running games on GPUs with the same amount of CU on the same clocks), Navi is 39% faster than Polaris according to that test that would mean 9.2TF on Navi would perform like 12.8TF on Polaris, so basically 3x faster than PS4 Pro's GPU and 2.13x faster than X1X GPU.
Stop using Vega 64, it is literally the worst benchmark for performance.You are also wrong here, because big navi is so inefficient, 9.3tf(on avarage) 5700xt is 1.14x faster than 12.58tf vega64 so it's like 14.34tf vega64, 9.2tf navi would be like 14tf vega 64.
I think Colbert has it.
5700XT is actually 19% faster in 1440p and 18% faster in 4K. So 1.18x 12.5TF is 14.75TF.You are also wrong here, because big navi is so inefficient, 9.3tf(on avarage) 5700xt is 1.14x faster than 12.58tf vega64 so it's like 14.34tf vega64, 9.2tf navi would be like 14tf vega 64.
I only own a Vega 56, not a 5700 (XT)
I mean the article that you posted here. At least I think that was you. That compared IPC of Navi to other architectures.
If we like big tflops number why not? ;) Comparing more reasonable to vega 56 9.2tf navi would be like 13.4tf vega 56Stop using Vega 64, it is literally the worst benchmark for performance.