Sapphire's Radeon RX 7800 XT Nitro+ is the quietest RX 7800 XT reviewed today, it's virtually inaudible at full load thanks to its powerful cooler. The card is also the best overclocker, it gained a spectacular 15% in real-life performance, which makes it nearly as fast as RTX 4070 Ti.
To me it seems like AMD should have a option where it uses Radeon Super Resolution and Fluid Motion Frames together at times when the scene is more relatively still in terms of motion and player interaction and then offset it with using Boost and Anti-Lag when the scene is more actively in motion. If they could strike a nice balance between the two it work really well. Reduce latency and boost frame rates when it's most needed and boost image quality when it isn't and when latency isn't of importance or as high importance.
I wonder if they could do something like that and then have some balancing slider strength controls over the 4 settings. Might even be able to swap boost and fluid motion frames as well which in certain instances could make more sense to swap emphasis prioritization.
I really think that combining Radeon Super Resolution and Fluid Motion Frames at the right opportunistic usage times would be beneficial to image quality and w/o any major downside. Using them at times when latency isn't a concern makes a lot of sense. Basically a nice image quality uplift with better frame animation at times where latency isn't a high priority. It's what Nvidia should've done in the first place with frame interpolation as a starting point.
Boost and Radeon Super Resolution would work rather nicely with post process like reshade or alternatively Boost and Fluid Motion Frames giving most post process and better frame animation. Having some tailored weighted control over the different features would be a nice consideration. It could even be configurable based on game profiles given what you might considered ideal could vary based on games involved and expectations around them. That's one of the biggest positives strengths of reshade you can easily adapt it to suit needs based on invidual game by game basis.
A RDNA3 CU, despite its similarities to a RDNA2 CU, has differences too and that's why a 60 CU 7800 XT is significantly faster than a 60 CU RX 6800 despite not clocking much higher.
Not a fan of how this time around Sapphire has ditched their dual heatsink(a dedicated heatsink for VRM and RAM + larger heatsink for GPU) cooling solution and easy to replace fans.
I've just purchased this one based on @W1zzard's teardown. The build quality of this one is top-notch and well worth the small 6% price premium over here.
The card seems fine. The amount of memory is spot on, and the bus is fine too. But the generations and thus the price is a bit off. I honestly think this should be RX 7700XT, and the 7700XT should be 7700. This makes sense even more, taking into account that quote: "AMD considers the RX 7800 XT to succeed the RX 5700 XT". So the price should be a good hundred bucks less. Not to mention 5700XT had $399 MSRP. But anyway, be it at least 75 bucks less, it'll be a good candidate.
Also, I wonder what the latency between chiplets is? It's not the biggest issue for me though. But there are games that are latency-sensitive. This is obvious that 7800XT is not a successor to 6800XT by both CU count and latency, but I doubt it's a true successor to 5700XT either, considering the last had a monolithic chip.
Video playback is still nuts. 44W for running a video. For people who watch YT for entertainment or especially tutorials, this is no go. The sad thing, despite all the fuss around the chiplets, and how AND were pushing it, there's almost no power efficiency added by this transition. Almost 50W less compared to 6800XT... Are they serious. This is within undervolt procedure area. And considering the performance gains are almost solely due to higher clocks... This is bad.
One thing I don't give a crap about is 120 AI accelerators. I know everyone is wetting their pants due to AI. But I'd better get more CU or L2 cash, instead of this thing. Or even simply without this area included, which would massively lower the price of GPU instantly.
I am against the notion of stuffing the enterprise stuff into consumer products. This does no useful service, but inflating the price, the way JHH does four years in row, since inception of RTX "holy-grail" by including the tensor cores, which have no use in "gaming" sector. The ones that require these Tensor and AI accelerators for their work, definitely should able to go and buy themself Professional cards.
Not a fan of how this time around Sapphire has ditched their dual heatsink(a dedicated heatsink for VRM and RAM + larger heatsink for GPU) cooling solution and easy to replace fans.
Indeed! I honestly thought Sapphire the best AMD AIB, but the decisions like this make to look elsewhere. Not only their most useful unique feature of easy-swapable fans for Nitro+ is gone, along with simplifying heatsink, turns Nitro+ into rather fat Pulse with rainbow puke. Fan cleaning here will be heck of a stunt, because some countries prohibit the card disassembly. And there's nothing "easy" in cleaning fans in Nitro+ since it requires of removal of the shroud completely. Tell me how this is any different to Pulse version.
The RGB crap being now forced, which I don't give a dang about. And it rises price much. The Pulse however, is not a non-RGB aternative, and it's still way too inferior, compared to Nitro+. Since not only the heatsink is weaker, but the bios options are missing too.
Sapphire push the artifical segmentation too strong.
Despite all the rant here, I still prefer Sapphire among others.
A RDNA3 CU, despite its similarities to a RDNA2 CU, has differences too and that's why a 60 CU 7800 XT is significantly faster than a 60 CU RX 6800 despite not clocking much higher.
14-17% based on resolution. In no way is it a significant generational upgrade.
Even less: 3-4% against 6800XT because AMD decided to name what should have been 7800 as 7800XT and thus people will compare it with 6800XT not 6800.
Tests have shown that RDNA3 is barely faster at the same clocks and CU counts compared to RDNA2.
14-17% based on resolution. In no way is it a significant generational upgrade.
Even less: 3-4% against 6800XT because AMD decided to name what should have been 7800 as 7800XT and thus people will compare it with 6800XT not 6800.
Tests have shown that RDNA3 is barely faster at the same clocks and CU counts compared to RDNA2.
Prices can be changed today. Changing chip sizes and CU counts is a years long process. So thanks but no thanks. I will continue to go by name/chip size and CU count when comparing generational advances.
I've just purchased this one based on @W1zzard's teardown. The build quality of this one is top-notch and well worth the small 6% price premium over here.
So 4% faster in raster than 6800XT while consuming 50W less power. That's actually better than I thought and with the same CU as the 6800, it's about 17% faster while consuming almost the same power. That sums up the RDNA3 arch pretty nicely, it's not groundbreaking but it's better than people who claim that there's barely any change from RDNA2 to 3.
Compared to the 4070, it's all around a better deal and overclocks better too. If only AMD priced it at $450 which would pretty much make the 4070 pointless.
14-17% based on resolution. In no way is it a significant generational upgrade.
Even less: 3-4% against 6800XT because AMD decided to name what should have been 7800 as 7800XT and thus people will compare it with 6800XT not 6800.
Tests have shown that RDNA3 is barely faster at the same clocks and CU counts compared to RDNA2.
IMO they released a 7700XTX and if they had of called it that they would have saved a lot of flak. Just like the 7900XT is the actual 7800XT. The 7700XT does raise the bar and about the only model they named correctly.
Surely you knew exactly what to expect? The specs were leaked months ago and confirmed several weeks ago. The 7800XT has 25% fewer cores and ROPs to work with, and (with the exception of the AMD reference-model 6950XT) the clock speeds are very close to equal at about 2500MHz boost.
In order to overcome a 25% compute deficit, RDNA3 would need 33% higher IPC than RDNA2, and we have nine months of reviews around the web proving that RDNA3 is somewhere close to zero, with a few specific examples giving measurable, but still only single-digit percentile gains.
The only variable that muddies the water is the much larger cache on the older RDNA2 cards. It's why the 6950XT scales to 4K slightly better than the 7800XT, but neither card is really a 4K card with 2023 AAA titles, and the scaling improvement is really small so that's a moot point IMO.
Not sure, were many people expecting it to be slower? I think most people were expecting it to be in the same ballpark with 17% fewer cores and 10-15% higher clocks with a little more bandwidth. Turns out they were right because we've know everything about RDNA3 since last year and it takes very little effort or intelligence to extrapolate that to a different core/clock configuration.
I think I said I'd be surprised if it was faster, but unsurprisingly it's the same speed, exactly as the numbers always suggested.
Anyway, my 7800XT Nitro+ has just been dumped on my desk by the courier, so I'm off to see how far I can undervolt it....
Blame AMD for their idiotic naming. If you want to see what the real gen over gen change is, compare it to something with the same number of CU (6800).
I think what you are trying to say is there is 0% increase in fps/dollar.
A man can dream
I guess it overblocks better than the 6800XT?
BTW amd also restricted the 6000 series from the antilag+
At the end of the day i don't think i'll be dumping my 6800xt for it, specially now that i know mine will support everything except antilag+ (i don't use av1)
