Show us some benchmarks already
Accurate.
I'm pretty f'in toasted on talking about it until we know everything about it. (For me at least) it's fun to speculate and try to understand what they may do with the product, but this has been excessive.
I mean, this is our hobby and it's what we (and I) do...but sheesh. This prolonged escapade has been tiring for everyone, I think (and yes I do admit to my part in contributing to that).
they're back to monolithic, used higher bandwidth VRAM modules and buffed up the CUs, as long as it gets perfs in the 7900XT-XTX gap for a decent enough price, I find that acceptable, no need for 4090 performance as long as it satisfies upper mid/high gaming reqs since it's what it's made for
VRAM isn't faster though...that's the whole (potential) problem. If you think these are faster will likely depend on how you measure performance. RT buffed probably ~30% (per cu).
...and then you have to go by things like actual memory bandwidth and/or if you overclock. The actual flop/buffer if you overclock 7900xt will exceed these cards. If you don't,
Well, then 12288*2631 (7900xtx)/ 8192*2970 = 1.33x. There's a good chance *in not compute/buffer/bw-limited scenarios* it, or higher clocked models could match or exceed it in RT.
This is about the difference between nVIDIA and Navi3 (performance per unit per clock) and odds are by doubling TMUs (or whatever TAUs they added) likley could/will make up the difference.
What matters at this point is pricing and availability.
Getting hold of an MSRP 5070Ti yesterday seemed like a lot of effort and a lot of luck, but we know that retailers have had 9070 series cards in stock for 6+ weeks at this point - so hopefully they won't be scalped the same way Nvidia's "$749" 5070Ti was.
I'm hoping for a $649 price on the stock-clocked XT. That's higher than the $500 price point AMD said they were targeting, and still a disappointment - but with the 5070 Ti currently going for $900, I think $649 would be an appealing alternative, provided it's genuinely at the 4080-tier raster performance and 4070Ti-tier RT performance.
I'm not so worried about the 304W TDP if the vanilla card uses 220W - it just means that the default clocks/voltages are being pushed far too hard for the XT and a bit of light undervolting and perhaps limiting the clocks to 2950MHz instead of >3.1GHz will get me 97% of the performance for a 20-30% power reduction.
I'm still hoping AMD is trying to surprise us with the prices we initially expected.
I still think everything points to it *used* to be ~400/480 (many accounts of this from CES plus leaked bulgarian prices converted), and we saw that one (Gigabyte?) model listed for $530. That makes sense.
Listen, if I'm wrong I'm wrong and that's fine...It's not cope, it's a very fair/realistic opinion...but they could do it, and they should do it. If they changed it, they changed it...but then it's just a worse-to-bad deal!
Again, I'm as tired of this discussion as everyone...I truly am.
Can't possibly be that much more, if it's a bigger chip than lets say Navi 32 then it's because it probably has more cache. It doesn't matter as the shader count itself says it probably wont have the same performance.
Cache is possible, but again it also could be related to both the increased texture capability (for ray-tracing) we know to be true (gpu-z) if not allowing for much higher clocks. Again, the later long-speculated.
Again, look at 5070's die size. It is also strangely large for it's configuration (similar to 4070 series but 80% of units). Does it clock 25% higher? I don't know. Maybe! (Probably, actually).
This is why people should not be surprised by 3400-3500 (or even potentially greater) clockspeeds. They *could* happen. The process is built for it. Do I *know* they'll do this? No.
304 seems weirdly specific, like, they could have just said 300 or 305 and we'd have read the delta as margin of error or something, not sure why they felt the need to give out such a precise number...
besides that, I also saw the news about XFX's 9070XT magnetic air and it's looking real fine to me ! can't wait to snag a white one for my build !
304 * 15%+ PL = 350W
Like i've speculated...pretty sure they're binning and power/clock locking these chips so they fit in very particular categories. Will wait for W1zard's analysis to confirm.
I'm very curious the clock potential of the chip. If the cache isn't increased, and the VRAM isn't increased, but it clocks well then this (XT) being the top product makes no sense. We shall see.
All I know is, don't (necessarily)
trust Frank. That's his job.
