adroc_thurston
Diamond Member
- Jul 2, 2023
- 7,756
- 10,471
- 106
More like they improved SE-level scaling with some trickery: https://patentscope.wipo.int/search/en/detail.jsf?docId=WO2025144455&_cid=P21-MCPWUB-70374-1Is SE scheduling that much harder than WGP scheduling in each SE to fatten them up that much?
Eh even with all dense fins clocks go up.Oh no it's a GPU.
I just can't see AMD releasing a gpu with TDP of ~800W. So either the hydrogen bomb is just a sketch in the AMD lunch room wall, or they have some nice IP that can reduce the joules per gpu-ops.That is the halo config I've been thinking about for some time, would need around 800W or so to hit those clocks I think.
I hope - for our sake - Nvidia and AMD come to an agreement to limit client graphics cards to 600W.I just can't see AMD releasing a gpu with TDP of ~800W. So either the hydrogen bomb is just a sketch in the AMD lunch room wall, or they have some nice IP that can reduce the joules per gpu-ops.
Big boy accelerators are going to 3kW a pop pretty soon.I just can't see AMD releasing a gpu with TDP of ~800W
well, PCIe AIC FF is the limiter, really.I hope - for our sake - Nvidia and AMD come to an agreement to limit client graphics cards to 600W.
The Venn diagram between AMDs shader cores and what defines a CPU core is getting closer to a circle.More like they improved SE-level scaling with some trickery: https://patentscope.wipo.int/search/en/detail.jsf?docId=WO2025144455&_cid=P21-MCPWUB-70374-1
Good 5090 air coolers can handle 800W right now without too much trouble.I just can't see AMD releasing a gpu with TDP of ~800W. So either the hydrogen bomb is just a sketch in the AMD lunch room wall, or they have some nice IP that can reduce the joules per gpu-ops.
I can see it, 600W hardlock with a 'secret' OC BIOS that voids warranty.I hope - for our sake - Nvidia and AMD come to an agreement to limit client graphics cards to 600W.
Pretty much, when do we get the babby version of OAM on desktop?well, PCIe AIC FF is the limiter, really.
We can call it the 0.8hp limit. 😂Or just do what the Japanese did with the 276hp limit, lie.
When ATX dies by the way of a bloodied hammer (i.e. never).Pretty much, when do we get the babby version of OAM on desktop?
see the issue here is dumping 800W into the case of some no-good gamer(tm) idiot.Good 5090 air coolers can handle 800W right now without too much trouble.
eyyy that's for chiplet graphics.More like they improved SE-level scaling with some trickery: https://patentscope.wipo.int/search/en/detail.jsf?docId=WO2025144455&_cid=P21-MCPWUB-70374-1
AMD loves the zoomzoom, but GPUs are throughput machines so 3-2 is unwarranted.Sticking to 2-2 Fin I see than
Have finflexAMD loves the zoomzoom, but GPUs are throughput machines so 3-2 is unwarranted.
This actually makes me wonder if poverty Medusa big cores would be 3-2 or 2-2.
No nanoflex there, d'oh.
FSR Redstone is public.But gamers buy upscaling technology and that wasn't outlined on the chart
Cuz it's public already lmao.Why not leak the important information that gamers need?
Oh it doesn't have DLSS? I think it'll be hard to sell.Cuz it's public already lmao.
FSR4 reception was so good they plastered it front and center on 9060XT marketing.Oh it doesn't have DLSS? I think it'll be hard to sell.
nope.
Well yeah. They've surrendered.
Just gave up.
Roadmaps are not set in stone.So there a halo RDNA5 after all? Weren't you guys saying it didn't exist?
Like.. for 7 months at least?
They could've changed yesterday.The roadmaps changed last week?
The halo may not be a "gaming" GPU. I don't think AMD can sell a $2000 gaming GPU . Its primary purpose could be something else. Something like the Radeon VIIThe roadmaps changed last week?
View attachment 127684
I'm not interested in "gotchas". I'm more interested in knowing what led to this change (assuming there was a change from late '24 roadmaps).
Yeah they can.I don't think AMD can sell a $2000 gaming GPU
