Monday, June 10th 2019
AMD Radeon RX 5700 XT Slide Leaked: Picture and Specs
Ahead of its official reveal at AMD's E3 2019 keynote scheduled for 3 PM (Pacific) later today, VideoCardz scored a key slide that spills the beans on AMD's next performance-segment graphics card, the Radeon RX 5700 XT. This card is based on the 7 nm "Navi 10" silicon, and is its "XT" (maxed-out) SKU. Its reference-design board design in the slide reveals a return to a lateral-blower type cooling solution that now has a prettier cooler shroud with silver ridges and Radeon logos on two sides, one of which is illuminated, with a possible RGB LED accent that runs along the top of the card.
The specifications revealed point to 40 compute units. Unless AMD changed the stream processor count per CU with the RDNA architecture from 64, this works out to 2,560 stream processors. When combined with a stellar engine boost frequency of up to 1905 MHz, the GPU has a compute throughput of 9.75 TFLOP/s, which is 37 percent higher than that of the RX 590, but 27 percent lower than the Radeon VII, and roughly similar to the RX Vega 56. The RX 5700 XT is armed with 8 GB of GDDR6 memory, although the slide won't mention memory clock-speeds or bandwidth. AMD may disclose pricing and availability in its keynote address later today.
Source:
VideoCardz
The specifications revealed point to 40 compute units. Unless AMD changed the stream processor count per CU with the RDNA architecture from 64, this works out to 2,560 stream processors. When combined with a stellar engine boost frequency of up to 1905 MHz, the GPU has a compute throughput of 9.75 TFLOP/s, which is 37 percent higher than that of the RX 590, but 27 percent lower than the Radeon VII, and roughly similar to the RX Vega 56. The RX 5700 XT is armed with 8 GB of GDDR6 memory, although the slide won't mention memory clock-speeds or bandwidth. AMD may disclose pricing and availability in its keynote address later today.
52 Comments on AMD Radeon RX 5700 XT Slide Leaked: Picture and Specs
"Cheaper and about as good" has proven to be completely useless for market share in the GPU market.
They might as well get more for what ever little number of cards they sell.
The market has decided that we are willing to pay for more expensive GPUs.
If it gets 122 FPS, it will be awesome.
Well, now we could see other benchmarks.
RX 5700 XT gets 53 FPS and same with GTX 1080 Ti. I only say that It is impressive.
Okay but i don't want it is same with R7. AMD why released R7 for $699?
All right, Why is RX 5700 XT worse than RTX 2070 in SoTR?
I wonder 1080p benchmarks. Is it better or worser than RTX 2070?
I'd love if AMD would not drop the price, and just have faster, similarly priced card.Because at times, when needed, 2070 ain't faster than 1080, but is "basically equivalent".
Makes perfect sense.It is notable, that significant lead comes in RTX enabled games.
So, either playing dirty and having RTX on, or, uh, heavily denoised crappy ray tracing also from AMD.AC:O in particular, in which 1070Ti stops on V64, yeah, absolutely the best case for sure.
We will see Nvidia VS AMD and in 2020 we will see Intel VS AMD VS Nvidia.
Leapfrogging right back into 2-3 year old performance. Navi has been pushed back so far that it is now upper midrange. And that is even with a competitor slowing down, because Turing really is Nvidia doing a very weak generational perf boost. Radeon VII was a letdown for showing us what 7nm could do - and Navi 10 is hardly breaking that trend. So we are left waiting for a Navi 20 that will again be behind the curve.
AC:O (yes, a greencampfuckamd optimized game, but 5070 XT beat 2070 in it) is notably missing.Uh, that's a nonsensical take on what's going on.
2080Ti is a 754mm^2 monstrosity.
Not having something in that area doesn't mean "being years behind".
2060/2070/2080 is a today's "right here and now" line-up, not "from years back". (I love how you refer nvidia's ever inflating price policies though)
2060 < V56 < 5070 (non-XT)
V64 < 2070 < 5070XT (40CUs)
VII < 2080
2080Ti - no fucks given, since underdogs do not have spare R&D money. (where is Volta, on which, as Huang said, NV spent billions to develop?)
Polaris cards wee roughly on par with Pascal perf/mem/die size wise.
Vega notably lagged behind.
Navi is restoring the parity.
It has node advantage, but it's also smaller chips.
The only aspect in which AMD is notably lagging behind, is power consumption.
Thing is, the only parity that is restored here is cost effectiveness. There is nothing new on offer so AMD is once more relegated to competing on price, and price alone. Not hard for Nvidia to steamroll right over.
BTW it seems we give an equal amount of f*cks about that 2080ti ;) It will be interesting to see what Nvidia will do with 7nm. And how much Navi 20 improves on 10 (because it still needs to).Don't compare benchmarks from different sources, it will skew your view on reality. W1zzard has a different test setup so they may be related, but they don't match. That answers your questions.
So, oh well, that's a card from the past. Forget it.No, there is perf/die size on a not super-expensive memory parity.
The "nothing new" argument is weird and can be applied to:
1650 (why do people even buy this piece of shit)
2060
2070
As that performance was available for that money years ago.
Even more cards can be called out, if I bring in ridiculously cheap V56/V64 cards.
And, hold on, I recall well OC-ed 980Ti are not that far from 1080, for the price.
With Navi 10, if it lacks aggressive pricing, we're stagnant.
2) 1080Ti is an interesting fluke, since you can't even buy it (so, a downgrade in availability, cool times, eh?)
Just chatting, not objecting, peace & harmony.I'm pretty sure 99% of people buying GPU's couldn't care less for $1000+ cards.
Once NV moves to 7nm (Samsung?) will probably be way past Xmas.
AMD could have "next" RDNA cards by that time (they kinda have to, since both Microsoft and Sony want that) with DXR, heavy "AI" denoising of that weird supernoisy "something" that one could use to mimick reflections/etc with significant performance penalty.
But even if they don't, haha, so what?
If NV will be able to "tier up" performance wise, if AMD won't, they will continue to play that "pricing tier up" game. 970 ($330), 1070 (spare me, don't even want to dig it up), 2070 ($500).
So, what about $599/$699 for 3070, hehe?
I'd thoroughly enjoy it.
I believe that Nvidia's Ampere will strong and cheaper than Turing. I expect that RTX 3070 will stronger than RTX 2080 for 400-450 Dollars, RTX 3060 will same with GTX 1080 Ti for 300-350 Dollars, RTX 3080 will beat RTX 2080 Ti with +%10 performance for 600-650 Dollars, RTX 3080 Ti will stronger than RTX 2080 Ti by %30-40 for 900 Dollars.
7NM has big density which is up 96.5(TSMC) or 95.3. It is 3.39 times Pascal's Density as well as it has got high clock speed. Think of the Ampere's power. Possibly they will make refresh.
If this is the low-to-mid end then I guess the high end should have 24-32GB RAM right? That's going to be good for simulator games.
I'm not getting anything below 16GB of VRAM, that's for sure.
Lol.
I hope I'll remember your post, once <whatever next gen> comes out of nVidia.
For a prospective buyer in my country this translates to the stagnation I am pointing out. And that is what happens if both players cannot keep moving up in performance - you get last year's performance resold to you at a premium - and AMD is guilty of that as well as Nvidia.