I personally don't think any should,w as joking, people are expecting too much, get dissapointed, and blame the designer. It is a console afterall.
Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]
- Thread starter BRiT
- Start date
- Status
I personally don't think any should,w as joking, people are expecting too much, get dissapointed, and blame the designer. It is a console afterall.
Metal_Spirit
Regular
Haha, wasn't meant to mislead, my post on GAF was headed with "THE DREAM SPECS FOR ME..".
And I'd also posted more rudimentary versions some time back on REE (I bounce between forums as I don't like to get stuck in echo chambers).
Never meant for it to be construed as a leak.
Well... You do dream with your feet on the ground. For a wet dream, it is quite dry, and not over exagerated.
You shure fooled me.
They were sent to work on Google Stadia instead.
$449 if exactly 10tf ?
AbsoluteBeginner
Regular
I just want to clarify for people who dont have Github docs, Oberon is almost certainly chip going into production and 2.0GHz is most definitely not "Pro boost". It clearly says "native", many times that is.
Given that it is regression test its only question if this is in fact 36CU Native or was version before it 36CU Native, and this one might not be. I am almost certain its 36CU, as other chips native do match with their actual hardware, but lets see. One thing is clear, it was not a stress or pro boost test.
Given that it is regression test its only question if this is in fact 36CU Native or was version before it 36CU Native, and this one might not be. I am almost certain its 36CU, as other chips native do match with their actual hardware, but lets see. One thing is clear, it was not a stress or pro boost test.
D
Deleted member 11852
Guest
It took AMD two days to have the files removed from github. Way too late.
Two days ? More like half a year. Komachi did warn them when he found the repo.It took AMD two days to have the files removed from github. Way too late.
D
Deleted member 11852
Guest
Yeah, but two days after the existence of this information became high conjecture for the public who knew how to use google. It took me maybe five minutes of googling to locate it. And I was half-asleep and hungover.
D
Deleted member 13524
Guest
Oberon might be the chip with the architecture going into production, though most definitely not the revision for production because it was fabbed in the beginning of 2019 at the latest.
36CU might be "full chip" only in the eyes of the people testing BC, as there's little reason to implement more than the amount present in a fully enabled PS4 Pro. Same with 256bit memory bus.
Assuming 2GHz is the absolute maximum that the final GPU can ever run at in the final console is also over conservative, again given the fact that the Oberon seen in the github leak was fabbed ~2 years before the console's release.
AbsoluteBeginner
Regular
Oberon A0 silicon was from May.
Everything before was defined as "pre silicon test". V dev kits went out in June, so I guess by then they had actual APU.
Other chips have regression tests as well, and native was actual native for all of them. BC1/2 are only found in Oberon and Arden/Sparkman. There is nothing there that indicates native at 2.0GHz is BC test, as it uses formulas related to Navi GPU.
D
Deleted member 13524
Guest
Source?
The earliest test in the doc AFAIK is from May.
That means the chip was definitely not fabbed in the same month as it had to go through QC, packaging on a substrate, then QC again, then soldered on a PCB, then QC again, then assembly, only then sent to full system testers.
Someone here claimed that a prototype system being tested in May most probably had its chip fabbed in early 2019 or even late 2018.
Devkits sent out 17 months before console release may not have final clocks.
Source?
2.0 ghz would be nice as I was thinking the real clock would be like a less aggressive 1.8. That would push them into 8TF territory and be a PR disaster. Imagine the xbox One X encroaching on your next gen effort in pure flops.
If they actually hit 9.2 or above that's healthy.
If they try and boost the SOC power, two main options. Enabling disabled CU's (if it's 40 CU chip) and boosting clock. The latter would seem difficult starting from such a high rate. Anyway, Xbox is likely to have even more headroom for clocks so it might not help if it causes a counter reaction from MS.
Then there's RAM, not sure how feasible things are there. Somewhere it was said the dev motherboards could only take 20GB absolute max, like that's some kind of hard limit on these consoles (effectively 16GB)?
If they actually hit 9.2 or above that's healthy.
If they try and boost the SOC power, two main options. Enabling disabled CU's (if it's 40 CU chip) and boosting clock. The latter would seem difficult starting from such a high rate. Anyway, Xbox is likely to have even more headroom for clocks so it might not help if it causes a counter reaction from MS.
Then there's RAM, not sure how feasible things are there. Somewhere it was said the dev motherboards could only take 20GB absolute max, like that's some kind of hard limit on these consoles (effectively 16GB)?
AbsoluteBeginner
Regular
Literally Github data. First SOC result (column) data is in May. Only column filled out for Oberon A0 before that date was "pre silicon". There are columns with theory/pre silicon/post silicon/SOC results. There are many incomplete regression tests in repo that are only partly filled out and you will note that they range from 28th of March 7th of July (I think).
Oberon A0 SOC results appear to be from May. As there are many uncompleted tests, it can be seen when certain colum was filled out. To make things more interesting is Flute benchmark, found in July, had GPU ID 13F9 in code string. 13F9 is Oberon A0.
And I agree.
Github. Renoir_Native, Navi10_Native show actual native specifications. Arden_Native shows 56CUs, VRS and RT so I am relatively certain its related to native specs, not BC.
I want to clarify that 36CU in native might be referring to Ariel Native and that Oberon is actually updated chip with more CUs, but from Github is relatively certain Native does not refer to BC boost mode or thermal test.
Alot of these regression tests are not done on actual hardware I think. There are many tests done as specified "sim" and pre silicon tests.
32GB GDDR6 clamshell is there for Sony. 40GB GDDR6 for MS.
Likely 16GB GDDR6 + 4GB DDR4, and 20GB GDDR6 respectively in the retail consoles.
I am thinking pre-silicon is the virtual silicon they put through simulation to find bugs. If Ariel is pre-silicon and not an earlier, different APU, then the Ariel and Oberon should be the same configurations.
anexanhume
Veteran
Undefined
It took AMD two days to have the files removed from github. Way too late.
It took them over 6 months to pull those files.
Having them pulled proves they are GUILTY
AbsoluteBeginner
Regular
Tbh while Komachi knew where files could be found, he didnt want to share anything related to them and back in August said its very sensitive.
...then _rogame came and leaked it all lol
The guy from AMD even posted his CV. So very few people were aware of its existance.
Edit
Also while I can understand AMD/Sony/MS requesting data to be removed from Github, how come Rogame deleted it next day from twitter? They contacted him as well?
...then _rogame came and leaked it all lol
The guy from AMD even posted his CV. So very few people were aware of its existance.
Edit
Also while I can understand AMD/Sony/MS requesting data to be removed from Github, how come Rogame deleted it next day from twitter? They contacted him as well?
Last edited:
- Status
Similar threads
- Locked
