• Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
Status
Not open for further replies.
Jan 20, 2019
10,681
the chart i posted above shows 40 CUs at 1.7 ghz giving us a 110w for the gpu. so yes, thats the ideal most efficient clock it seems for the 7nm navi. we will see what happens with 56 CUs but it should be a linear increase in tdp.

2.0ghz consumes 2x more power at 210w. why would sony go narrow and fast only to end up with a 30% less powerfull console that consumes almost 50% more power?

If the leak is true, then Mark fucked up? Why whould he go that route...
 

III-V

Member
Oct 25, 2017
18,827
I'm not really throwing anything out.

The three test modes inform on how the part in question is being tested. It still doesn't fundamentally tell us anything about what I said being tested.

At most it proves a relationship between the Oberon codename and PlayStation. But that's about all it gives us.



I never made this comparison.

I'm merely stating that just as when we saw pastebin posts with obvious red flags and inconsistencies, we treated them with a level of scrutiny, we should also employ a similar level of scrutiny here.



I think some are doing this. But I think other are just innocently placing too much stock in the Github leak being legit internal AMD testing data, while failing to ask the important questions about what the info. is, what it means, while acknowledging that we just don't have answers for these questions just yet.



Then on this point we agree, good sir.
You asked the question of how do we know the Oberon data wasn't copy-pasted from a 5700. The back-compat and references to Oberon-silicon being distinct from Ariel. Well that's how. I am not claiming this is the final PS5 GPU or anything of that sort.
 

BreakAtmo

Member
Nov 12, 2017
12,805
Australia
I'm not really throwing anything out.

The three test modes inform on how the part in question is being tested. It still doesn't fundamentally tell us anything about what I said being tested.

At most it proves a relationship between the Oberon codename and PlayStation. But that's about all it gives us.



I never made this comparison.

I'm merely stating that just as when we saw pastebin posts with obvious red flags and inconsistencies, we treated them with a level of scrutiny, we should also employ a similar level of scrutiny here.



I think some are doing this. But I think other are just innocently placing too much stock in the Github leak being legit internal AMD testing data, while failing to ask the important questions about what the info. is, what it means, while acknowledging that we just don't have answers for these questions just yet.



Then on this point we agree, good sir.

This really is going to be a long fucking couple of months. Klee's confirmation of the PS Meeting being in February is very "Thanks, I hate it" right now.

God, it would be nice if they forked over the PS5 spec sheet at CES. Then said, "and if you want to see how these numbers translate into games, join us in February".
 

Sunlight

Member
Apr 22, 2019
375
It's an issue with the litho process and the masks are not compatible and they are very expensive to create. It is not tied to the Arch in that same way. N7P does not require any re-work to the masks.

Yes I agree 7nm+ is more expensive.


But if we consider the accuracy of the leaks then the best guess of Arden size is 345~355 mm2.
That indicates 7nm+.

Yes, but later on down line.

I feel the design libraries are different between 7nm and 7nm+, starting out development on the former and moving to the latter midway through would lead to significant rework of the engineering design, delays and a very real impact on the overall project schedule deadline, I.e. launch date.

For any console to be on 7nm+ they would need to have been designed for it from the outset, which is something I AMD many more here consider unlikely. Not impossible, but a slim likelihood.
OBR already has slow verification process, and it still changed version and re-tested in Nov. According to the leaker Acuariusz.

That's an indication of using different node and causing problems.
 

III-V

Member
Oct 25, 2017
18,827
Yes I agree 7nm+ is more expensive.


But if we consider the accuracy of the leaks then the best guess of Arden size is 345~355 mm2.
That indicates 7nm+.


OBR already has slow verification process, and it still changed version and re-tested in Nov. According to the leaker Acuariusz.

That's an indication of using different node and causing problems.
If we use 7nm 64CU*5mm2 each we hit 320 mm2 just for iGPU. If we add in everything else that makes sense, unless I am missing something.
 

Paragraf

Member
May 31, 2019
532
Russia
If Lockhart (Xbox Series S) is 4TFs, imo wouldn't any impact from a 3TF gap between Series X and PS5 be sort of mitigated? All Series X games would have to be designed around Lockhart anyway, and all multiplats in general, so would the consumer really see a difference?
Yes, just like you can notice the difference after changing resolution or other settings in any PC game for the last 20 years.
 

AegonSnake

Banned
Oct 25, 2017
9,566
"smart" design ;d or just catching up after 1 year of delay
lol there is nothing smart about this. we are all arm chair engineers here and i challenge you to find one poster here who would design a console 30% weaker that consumes 50% more power. and since it consumes 50% more power it would likely cost the same or even more.

btw, it wont just be 30% weaker if the AMD patent is right and the rt stuff is built into the TMUs. TMUs scale with CUs. the more CUs you have, the more TMUs. it will be far worse when it comes to ray tracing.
 

androvsky

Member
Oct 27, 2017
3,498
you will have to move the sweet spot much higher to get 2ghz to be 'ok'. and there is no indication that its 7nm+.

ysplb2a3f6b31.jpg


So sweet spot for 40 CUs is 1700mhz. To get 10% more performance your tdp is going up almost 100% from 110w to 210w.

it makes ZERO sense. The only explanation for such a shitty design is what Doctor Avatar just gave... i.e., sony fucked up in ways no one can imagine. you would need to have some of the dumbest interns designing this console. Cerny is going around talking about this fastest ssd of all time, a dedicated 3d audio chip which is most likely the psvr 2.0 processor chip bolted on the board and fancy new controller features. why would they try to save $15-20 on the silicon when they are spending extra on dedicated processors and fastest ssds known to man?
You're designing the console in hindsight, of course any intern can design better based on final, tested silicon. Don't forget there were rumors that the Navi thermals were much worse than expected. It's also worth keeping in mind that the thermal issues were discovered in time to make some changes, since we're seeing these test results that got filtered down to an intern six months ago.

I don't know how easy the changes would be to make, either try to fix the fundamental thermal issue (might not succeed), go to a smaller process (many unknowns, might not help much), or just slap a bunch of extra CUs on and drop the clock (might cause BC problems). Or some combination of the above.
 

SONofZARDOZ

Member
Nov 26, 2019
32
I don't expect the PS5 to be that much weaker than XSX. The test info is interesting but as many others have said, even if it's real it doesn't tell the whole story. This is going to be a looooong ass wait for these things to be fully unveiled but it's going to be awesome once they finally are.
 
Status
Not open for further replies.