Next-gen PS5 and next Xbox speculation launch thread - MY ANACONDA DON'T WANT NONE
- Thread starter Phoenix Splash
- Start date
- Status
Memory controllers can be changed for specific customers. MS for example went with custom route and used DDR3 memory controller for base Xbox One.
AMD can use any kind of modules their customers want.
Trusted journalists verifying using multiple trusted sources.
he means that Ubi developers will leak to jason or some other journalist the details.
then again,
dont expect nearly all those 14k developers to work on the dev kits lol.
until now even in the first party studios there were only select people with access to dev kits and information about them that they were not allowed to share with the rest of the team, i assume that will roughly still be the case with third party developers.
depends if the game is designed for anaconda and downscaled to lockhart or the other way around. i could see a scenario where the game is designed to be 1800 p on anaconda and runs 900p on lockhart.
Yeah, there's also that. But I think that would be in years to come.
DF also noticed it:
"If this is an engineering sample board, a potent power delivery set-up like this may simply be in place to put the new silicon through its paces. However, it is a touch concerning if this is an AMD reference board aimed at the consumer - as the configuration suggests that a standard 'blower' design with a single fan may be in place as a cooling solution, something that rarely works out well for Radeon hardware."
"If this is an engineering sample board, a potent power delivery set-up like this may simply be in place to put the new silicon through its paces. However, it is a touch concerning if this is an AMD reference board aimed at the consumer - as the configuration suggests that a standard 'blower' design with a single fan may be in place as a cooling solution, something that rarely works out well for Radeon hardware."
Lockhart will have at least equals or more TF than Xbox one x but i suspect little bit higher
4tf lockhart in a 2020 is a joke I don't know how people can trust this
Interesting patent from AMD:
HYBRID RENDER WITH DEFERRED PRIMITIVE BATCH BINNING
http://www.freepatentsonline.com/y2019/0122417.html
Looks like they're largely repeating buildzoid's analysis without giving him credit.
FWIW, I'm betting the analysis claiming 3072 SP is lowballing the CU count. It's probably closer to the 3584 number indicated by Jim (AdoredTV's) numbers, meaning 56 CUs. Vega VII was a shrink to 7nm and not optimized to it, and I wouldn't be surprised if the CU size shrunk given the design goals and supposed changes to the microcode to revert some things to the way Polaris did them.
HYBRID RENDER WITH DEFERRED PRIMITIVE BATCH BINNING
http://www.freepatentsonline.com/y2019/0122417.html
Looks like they're largely repeating buildzoid's analysis without giving him credit.
FWIW, I'm betting the analysis claiming 3072 SP is lowballing the CU count. It's probably closer to the 3584 number indicated by Jim (AdoredTV's) numbers, meaning 56 CUs. Vega VII was a shrink to 7nm and not optimized to it, and I wouldn't be surprised if the CU size shrunk given the design goals and supposed changes to the microcode to revert some things to the way Polaris did them.
Last edited:
Native driver support just moves the s/w part which translate DXR calls into driver/hardware calls from the emulation layer into the driver. DXR is in its essence DX compute. Thus you can either translate it into DX compute via some emulation layer (what DXR Fallback Layer does) or handle this directly inside the driver, mapping DXR API calls to your GPU compute capabilities. The second option is better because a) you don't have an emulation layer somewhat slowing things down for you and b) you can map DXR calls to your GPU h/w better because you're the one who made the h/w. But both are "h/w supported raytracing" since GPU compute is something which is running on a dedicated h/w - GPU.
This is why I've said that Cerny's notion of RT support in PS5 doesn't mean anything. PS4 "support" RT right now, it's just that with its GPUs it can't make real widespread use of RT due to performance deficit. And unless PS5 GPU will have some hardware in addition to pure GCN GPU compute (async or whatever) I don't see this changing with PS5 much.
It wasn't.
There's no indication that this is the case. However it's worth remembering that next gen consoles are highly unlikely to use the same CLN7FF node which Vega 20 is using. Navi 10 (or whatever will launch this year) likely will though.
Last edited:
Yea that's a more useful breakdown - not sure if there's any way to make terminology stick, but it helps conversations to call out the differences.
I'd tend to agree, I don't expect a ton of fixed function work there(computing whole intersection tests like RT core - I doubt it). But based on the work AMD/Sony did for things like VR facing features and CB acceleration etc. - I do expect 'some' specific hw-customizations that will help the process.
Perhaps you can elaborate on what you think exactly Vega VII is and what node consoles will be on, then?
Don't they still have it wrong about what Arcturus is? I thought it'd had been explained months ago it is a product, not a microarchitecture?
Indeed. Just like DF is still listing a 4TF Lockhart.
Last edited:
They want price sensitive gamers to be gamepass gamers. Why would ms give two fucks about retail and used games?
To keep consumers happy and not be seen as forcing people into buying games a certain way like what happened in 2013
Who really knows. Maybe I am making a mistake compare it to the Xbox One X. But it would be weird if Lockhart would have less peak performance than the One X. Would be a marketing nightmare.
If Lockhart is $299-$399, with ssd and premium cpu...well, how many TF gpu do you think is reasonable given XB SAD is $250.
Navi was always expected to be GCN. I think that WCCF was the only site which were saying that it isn't. 64CU limit is not a limitation of GCN either, more a limitation of their current frontend which can be redesigned while still using GCN architecture for multiprocessors array.
Consoles assuming they'll go into full production at the end of 2020 are very likely to use either the N7+ process (the one with EUV usage) or the recently announced N6 process - which is more or less a shrink of N7.
Vega 20 was designed specifically for N7 process from the start. There's nothing "shrank" about it.
Consoles assuming they'll go into full production at the end of 2020 are very likely to use either the N7+ process (the one with EUV usage) or the recently announced N6 process - which is more or less a shrink of N7.
That is something i agree about:
"My xbox one x is 6 teraflops, why cant [insert next gen only game here] run on my console when it runs on a 4 teraflops console!"
They can do that by making gamepass a gentle breeze blowing through your bank account every month and having all first party new games on there day one. Yes there will be some who complain, but it wont be like 2013, esp with eXpensive box as an option.
You know this is all like putting a blind man in a dark room.
Considering GCN debuted over 6 years ago, I'd say it's not, no.
They wont be using TF to push cheapbox at all, so it doesn't matter as far as marketing goes. It is a cheap option to play next gen games with new tech like ssd and zen, and they will pump out a bunch of anaconda footage for dat halo effect. + gamepass! Keep your kids quiet for the cost of a Netflix sub! They've got lots of sales hooks besides more power.
Sure we'll have to wait and see but with a much newer GPU, massively better CPU, same amount (12GB) but faster RAM as 1X and secret sauce, it could perform great at 1080P even if paper specs are 4TF. Also Microsoft might not highlight TF numbers as mentioned by joe-zazen.
He's basing this purely on the idea that Sony wants 399 for the price
Many people disagree with that premise
GPU is not the only measure of performance.
A 4tflop gpu, 16-24gddr6 and a 8core zen 2 would be more caperble then the X1X.
However overhead is needed to guarantee 1080p, so I would expect at least 4.8tflops if anaconda is 12tf.
Lockhart needs to be at least 40% of anacondas gpu to have 1080p in the majority of games.
A 4tflop gpu, 16-24gddr6 and a 8core zen 2 would be more caperble then the X1X.
However overhead is needed to guarantee 1080p, so I would expect at least 4.8tflops if anaconda is 12tf.
Lockhart needs to be at least 40% of anacondas gpu to have 1080p in the majority of games.
I dunno, the 6TF X1X runs some current gen games at 1440p, trying to push next gen visuals on a 5TF GPU sounds like a nightmare even at 1080p.
Very well then Sony can counter with ps4 and psnow. That would be cheaper and psnow has a lot more traffic doesn't it?
The problem I see with this is that MS first party titles are not the best quality. They're "good enough for gamepass", pay a couple of dollars a month for some b tier timewasters.
I just can't see the big third party titles showing up on gamepass and cannibalising their full price sales for cents on the dollar. So you're left with the big MS franchises that the mass market still aren't interested in.
Even on resetera polls have shown it's games that sell hardware. The mass market punters are even less likely to buy in now for stuff coming down the pipe later (or not at all!) from recent acquisitions.
PSNow has been around for years and while it continues to grow, it's still relatively small. Microsoft and Google coming really hard at game streaming in the next year is really going to test how successful PSNow actually is.
if a 6tf gpu cant do 1080p nextgen, no way 12 does 4k. while not everything a a liner scale, resalusion is. to display the same 1080p image at 4k requires 4x the gpu power. a 6tfp gpu will not only be fine for 1080p, it will be more room for devs to work with, than a 12 tfp gpu at 4k.
If PS5/Ana is in the 12tflop range, then yes next gen games doing 1440p will need further cutbacks on a 5tfllop, probably 900p and some reduced settings.
I would imagine developing lockhart would be a similar process to the development of the X1X but reversed.
1st party and maybe some 3rd parties will have been developing for anaconda and with that data they will be able to determine what performance is required for a 1080p console.
It won't be perfect, like the the X1X is not a perfect 4k machine, but the results were still very good.
Also 1080p PC gamers will still be a thing, even if lockhart was not a thing, PC gamers who play at 1080p would need far less caperble GPU then a console pushing 1800p - 2160p resolutions.
- Status