• 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.
  • We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
Status
Not open for further replies.

III-V

Member
Oct 25, 2017
18,827
Arden tests in the github leaks? Easy to understand.

The intern was given info to do recessive tests on. If you look there's no actual tests in Arden and Sparkman, because they dont have real testing results yet, only the theoretical value.

They only have the cmodel.
I am not at my desk at the moment but I sort of remember some pass/fail testing on Arden or Sparkman
 
Jun 23, 2019
6,446
I absolutely think this is the case. I think it's why Komachi stopped posting consoles related stuff and Rogame is learning the same lesson the hard way. No one wants to be the first person to say anything.

Absolutely. I posted this last night:

lol Come on now. A rando tweets information from sensitive internal testing documentation done by AMD that was publicly available due to the work of a negligent intern. 2 days after the info is blasted all over Twitter, said rando suddenly deletes all tweets with information pertaining to the documentation. A few hours later, the repository where rando was getting information is deleted from GitHub. The only ones with the access to delete said repository would be the original entity that posted it (AMD). It's not hard to put 2 and 2 together here. Believe what you want though.

I think Sony has something huge brewing and secrecy is of the utmost importance right now. Just like MS kept Series X under wraps for TGA? Sony is about to respond.
 

Fabtacular

Member
Jul 11, 2019
4,244
I'm having trouble understanding the implications of the events of the past 90 minutes or so.

Would someone mind summarizing where we're at now?
 

Expy

Member
Oct 26, 2017
9,865
so many mistakes.
- intern mistakenly makes it a public directory.
- intern mistakenly pulls in oberon results in the arden folder
- intern mistakenly pulls in rt and vrs results from the arden results into the oberon folder
- intern mistakenly creates a folder named arden for a ps5 gpu.

worst intern ever lmao
I mean, it's their personal repo, not an AMD repo. You would think these tests would go to a corporately held GitHub account and not his own personal one that has literally almost nothing else in ib and is new as of a year ago.

You don't accidentally `git push personal master`. They knew they were pushing to their personal repo, and they can clearly tell it's a public one.

They've since removed the repo or made it private. But it's too late. That person is in for a lot of hurt @ AMD.
 

AegonSnake

Banned
Oct 25, 2017
9,566
How is literally nobody able to confirm the tf of PS5.
this is a good question. the anaconda 12 tflops number has been out for almost a year. only benji mentioned the 13 tflops number and even he deleted it asap.

maybe it is 8 tflops and they dont want to incur the wrath of disgruntled sony fans.

klee knows the tflops and he said nothing has changed. its 13 tflops.
 

Hey Please

Avenger
Oct 31, 2017
22,824
Not America
How is literally nobody able to confirm the tf of PS5.

I don't think any insider will "confirm" anything let alone the radioactive TF figure.

If there is water to the whole 1.8GHz core clock for GPU holds any water for the final package, the worst case scenario with 36 CU active is just about 8.3 TF. If as previously mentioned, (presumably) GCN dev kit's (rumoured) 12.9TF were worth their salt then a RDNA card with a TF value of about 10.3 would not only do the same job, but exceed it due to other features like VRS.

So, with half a pint of salt- Worst case is ~8.3TF and personally ideal (because it breaks the 10TF mark making brochure worthy) is ~10.2 to 10.3 TF RDNA on a GPU featuring 44 to 48 CU active.

Now, how does this fit into the early rumours of 300mm2 die space... well others are better prepped with that knowledge.
 

Betty

Banned
Oct 25, 2017
17,604
this is a good question. the anaconda 12 tflops number has been out for almost a year. only benji mentioned the 13 tflops number and even he deleted it asap.

maybe it is 8 tflops and they dont want to incur the wrath of disgruntled sony fans.

klee knows the tflops and he said nothing has changed. its 13 tflops.

Oh did he? A new hope it is then.
 

UraMallas

Member
Nov 1, 2017
18,926
United States
Just spitballing, but hmqgg floated the idea of Sparkman being Xcloud related rather than Lockhart (again he was just speculating). If Sparkman's only BC points to Xbox One S, would that mean that the 2 modes for Xcloud were Series X and One S? Does it make any sense for that to be case (since Xcloud is currently One S based)? Can't imagine it would but just a thought.

It would also explain the 56CU Sparkman.

I would say that makes sense but what do I know. If Sparkman is essentially the next gen Series X and the current Xbox One S combined into one, you could start swapping out the One S blades for the Sparkman blades when Series X hits, yes? The lower-tiered Lockhart would be able to stream Series X games with Sparkman/Series X specs through xCloud?
 

Dashful

Community Resettler
Member
Oct 25, 2017
2,401
Canada
Sony also have to test there BC which is what most of these test look like.

I was told last night or the night before that it couldn't be just BC. But what if if it was PS4 BC, PS4 Pro BC (for games patched for the pro), and a PS5 patched PS4 games mode?

With that patent we discussed earlier, the 36 CU would make sense for testing purposes, whatever other CUs wouldn't need to be there for these tests and in the final gpu/apu they would reduce down to 36 CU when you play a PS4 game using that patent technique.

Probably still not it. But I had to write it out.
 

gundamkyoukai

Member
Oct 25, 2017
21,136
this is a good question. the anaconda 12 tflops number has been out for almost a year. only benji mentioned the 13 tflops number and even he deleted it asap.

maybe it is 8 tflops and they dont want to incur the wrath of disgruntled sony fans.

klee knows the tflops and he said nothing has changed. its 13 tflops.

Klee was the only one to straight up give us a figure for PS5 which was 10+
Other people were aiming etc etc
hmqgg said he has Sony info but it don't look like he want to share .
As you can see Klee is the only one that do that on the Sony side.
 

Deleted member 43

Account closed at user request
Banned
Oct 24, 2017
9,271
I'm glad to see we've moved pass thinking the Github leak was confirmation of anything.
 
Last edited:

Bosch

Banned
May 15, 2019
3,680
DukeBlueBall can we hear your thoughts on hmqgg's assertion that XSX is RDNA1 (7nm)? That's going to be big...
It can't be different. We have nothing on 7nm+ and the node is not mature enough to have consoles production on April or May.

And this imply if they are pushing 11-12 tflops this console will consume more than 300w like df said.
 

Hey Please

Avenger
Oct 31, 2017
22,824
Not America
DukeBlueBall can we hear your thoughts on hmqgg's assertion that XSX is RDNA1 (7nm)? That's going to be big...

Technically speaking (and upon recalling previous events), Dr. Lisa Su presented slide shows of next gen console gaming comprising of 7nm Navi (RDNA) tech when unveiling Navi and latest Ryzen processors earlier this year. Next gen consoles were, iirc, never rumoured to be on the 7nm+ (upon which RDNA 2 is based). And so.... when did people start assuming XseX (or PS5) would be using RDNA 2 and not a customized version of RDNA?
 

Dekim

Member
Oct 28, 2017
4,301
All this info/drama over the past hour has told me is that almost no one knows anything, and those who do know something aren't talking.
 

gundamkyoukai

Member
Oct 25, 2017
21,136
Technically speaking (and upon recalling previous events), Dr. Lisa Su presented slide shows of next gen console gaming comprising of 7nm Navi (RDNA) tech when unveiling Navi and latest Ryzen processors earlier this year. Next gen consoles were, iirc, never rumoured to be on the 7nm+ (upon which RDNA 2 is based). And so.... when did people start assuming XseX (or PS5) would be using RDNA 2 and not a customized version of RDNA?

People did not think that way for Sny from what i seen .
It when MS start using RT and VSR that some say MS using RDNA2.
It always made the most sense that both using custom parts that will pick and choose stuff.
 

Noctis

Member
Oct 25, 2017
11,372
New York City
Round and around and around we go
tt7aGMQ.gif
 

Betelgeuse

Member
Nov 2, 2017
2,941
I'm glad to see we're moved pass thinking the Github leak was confirmation of anything.
There it is.

The preceding discussion that took that leak as incontrovertible fact is really funny.

To be clear - I certainly appreciate that the leak was far more interesting and of greater potential value than random Pastebin bullshit, but it never, ever remotely comported with the things you, Schrier, Warren, and Klee have said. When you have that mismatch, something is amiss.
 

III-V

Member
Oct 25, 2017
18,827
It can't be different. We have nothing on 7nm+ and the node is not mature enough to have consoles production on April or May.

And this imply if they are pushing 11-12 tflops this console will consume more than 300w like df said.
Yeah, going to be tough hitting 12 TF on RDNA1 it would seem, unless they moved to 7NP.
 

Betty

Banned
Oct 25, 2017
17,604
I don't think any insider will "confirm" anything let alone the radioactive TF figure.

If there is water to the whole 1.8GHz core clock for GPU holds any water for the final package, the worst case scenario with 36 CU active is just about 8.3 TF. If as previously mentioned, (presumably) GCN dev kit's (rumoured) 12.9TF were worth their salt then a RDNA card with a TF value of about 10.3 would not only do the same job, but exceed it due to other features like VRS.

So, with half a pint of salt- Worst case is ~8.3TF and personally ideal (because it breaks the 10TF mark making brochure worthy) is ~10.2 to 10.3 TF RDNA on a GPU featuring 44 to 48 CU active.

Now, how does this fit into the early rumours of 300mm2 die space... well others are better prepped with that knowledge.

Well even that worst case scenario doesn't sound so bad to me.
 

Deleted member 23046

Account closed at user request
Banned
Oct 28, 2017
6,876
Of course we'll have to wait for more info to be certain, but this explanation makes absolutely everything fit perfectly. The figures that were given at GDC. The figures we've been getting since then. Hell, the devkit shown off to Wired the first time was literally referred to as a low-speed devkit, lol. It explains the lack of RT and VRS in the Github leak too.

I can't think of any aspect that this theory doesn't explain.
Less than three hours after your message, we now have two insiders hints for the price of one. Enjoy the ride 🎢
 
Status
Not open for further replies.