As far the rumors goes, it will be on TSMC n3X node.

Will clock as high as 3.7 ghz.

This could potentially BTFO the RTX 5090, with less power consumption.

Does nvidia have anything in their sleeves?

>>105632487 (OP)
Just put out another named feature that requires renderer support and a nvidia card, watch engine developers rush to support it in exchange for money, and let AMD be stuck playing catch up again.

>>105632487 (OP)
>nvidia have anything in their sleeves
yes, investor money that they're looting from the company

9070XT uses like 330W, 9080XT will use what, 500W?

>>105632487 (OP)
Ray tracing performance?

>>105632764
Let's be honest, it's not like the 80 or 90 series are sipping power.

>leak
let me guess, it's from THAT guy that never gets anything right

>>105632487 (OP)
>rumors
>could potentially
Lol

>>105632487 (OP)
Im sure this gen amd will actually start competing for realz

Didn’t and last big card sell out fast and then they raised the prices on it?

They are as bad as nvidia don’t get your hopes up

>>105632797
No indeed, the power draw for both brands has been pretty bad for the current and past gen of gpus

If it can't crank out AI porn, then the higher memory capacity means fuckall.

AMDrones are mentally ill
It would be 100% in Radeon's playbook to screw the small amount of customers they have by releasing a more powerful GPU months after their previous one the overhyped and still sold like shit lol
AMDrones will call it a good thing lmao
AMDrones literally watch MLID videos and believe the things he said
Remember when he said 8GB 9060 XT was cancelled? (AMD CANCEL RX 9060 XT 8GB! (Leak) | RTX 5060 Ti = RX 6500 XT (Review Analysis))
lol

AMDrones gonna drone. This GPU market ended before it even started.

Nvidia doesn't care about AMD

>>105634119
Your mom is an Nvidia employee and she was trash talking amd so hard last night. Was so hot

>>105632937
this is how salty nvidiots sound like

>>105632487 (OP)
>nvidia forces devs to use path tracing only
rip amd

>>105632487 (OP)
How will it cope? By selling millions of units while amd sells none

>>105632487 (OP)
by reaching 98% market share before the third quarter

>>105632487 (OP)
i'm pissing and shidding myself in anticipation

>>105632487 (OP)
two more weeks amdsisters!

>>105632889
>If it can't crank out AI porn, then the higher memory capacity means fuckall
this
i just built a new computer after having my i7-8700k/1080ti build for ~8 years. I went with a used 3090 JUST for the AI

>>105632487 (OP)
> once
if

>>105632937
jesus, what an embarrassing meltdown

>>105632937
I've never seen that gif before. You are a paid shill.

>>105632889
ROCm works great for AI with my 7900 xtx, 32gb would be sweet

I'm a RTX5090 paypig but I'll unironically buy a UDNA halo product if it comes close. Fuck Nvidia's drivers. My 4090 was fine but the 5090 has sucked shit.

>>105632937
How long are you going to keep shilling for you fucking faggot hope the money was worth it for your soul

>>105637216
use the studio driver FAGGOT

>>105637160
ROCm is slow as fuck

>>105632487 (OP)
There's no reason to buy the 9090 XT if it has less VRAM than the 5090 or 9080 XT

>>105632487 (OP)
>mlid
like how this guy gets a pass here despite getting shit wrong all the time and never admitting it while calling adored an amdrone for getting one or two predictions wrong and admitting fault later

Honestly I'd be a little surprised, but not that surprised. The fact that they have a 9060 with 16gb shows me they care a lot more about VRAM than Nvidia does.

>>105632487 (OP)
>AMD: we're not going to compete on flagship
>RUMOOR: O M G THEY'RE GONNA COMPETE ON FLAGSHIP

>>105637160
ROCm might as well not exist, and ZLUDA is fucking shit

>>105632487 (OP)
Not going to happen, the largest RDNA4 chip will be always be 9070XT. Anything larger will be an UDNA chip geared for computer general/ML sold as instincts.
The simple truth is that gayming GPUs aren't hot ticket items for both AMD and Nvidia they rather allocate their fab space with TSMC to make more profitable compute compute/ML chips.

AMD do not have a 5090 competitor and never will

>>105632797
The 5080 is actually pretty efficient. Its TDP is 360W but it mostly sticks around 250 in my experience.

>>105632487 (OP)
The only way this product would make any sense is if its a dual gpu card.

>>105632487 (OP)
>rumors
>will
>could
>potentially

>>105632487 (OP)
Intel will drop their Battlemage GPUs too and tbqh I'd rather get Intel

>>105632487 (OP)
Moore's Law is Dead has literally never got AMD GPU leak right.

>>105632487 (OP)
ROCm support? 32GB is useless for gayming and the only use is AI.