AMD Radeon RX 7900 XTX and XT Review: Shooting for the Top

RDNA 3 dons a halo for its coming out party

AMD Radeon RX 7900 XTX and XT
(Image: © Tom's Hardware)

Tom's Hardware Verdict

The RX 7900 XTX and XT sport AMD's latest and greatest RDNA 3 architecture, with impressive rasterization performance. Ray tracing remains a second class citizen, and while faster than the previous generation, AMD can't touch Nvidia's in DXR games — and that's not even counting DLSS, which continues to see widespread use.

Pros

  • +

    Fastest and second fastest AMD GPUs

  • +

    Excellent efficiency gains

  • +

    Innovative chiplet design

Cons

  • -

    Chiplets don't actually improve performance (and may hurt it)

  • -

    7900 XT costs too much relative to the 7900 XTX

  • -

    Weaker ray tracing and AI performance

Why you can trust Tom's Hardware Our expert reviewers spend hours testing and comparing products and services so you can choose the best for you. Find out more about how we test.

We've tested the AMD Radeon RX 7900 XTX and Radeon RX 7900 XT, and we have the full results ready for your enjoyment — one day before the official launch party. AMD's latest and greatest RDNA 3 architecture and the RX 7900-Series graphics cards are all set to party with Nvidia's Ada Lovelace architecture and the GeForce RTX 4080 as they vie for a spot on our list of the best graphics cards.

One thing AMD won't do: Take down the GeForce RTX 4090 that sits at the top of our GPU benchmarks hierarchy. AMD has said that it doesn't feel the need to compete directly against a $1,600 (or more!) graphics card, but several senior people at AMD also indicate that Nvidia's AD102 chip was "bigger than expected" and basically out of reach.

What may also be out of reach for many of our readers are the new AMD RX 7900-series cards. While they cost less than Nvidia's RTX 4090 and 4080, with prices starting at $899 for the slower — and frankly less desirable — RX 7900 XT, these are clearly not going after the mainstream gamer market. That task will likely fall to the future Navi 32 / RX 7700-series cards (or maybe 7800-series). For now, if you want AMD's fastest ever consumer graphics card, be prepared to fork over a wad of cash.

We've already done an in-depth look at the RDNA 3 architecture and a preview of the cards, so start with those articles if you want to get up to speed. With actual hardware in hand and a bevy of benchmarks under our belts, that's the main event today. But we do have some additional thoughts, and we'll start as always with the specifications of AMD's latest cards, with Nvidia and some previous generation GPUs for comparison.
 

Swipe to scroll horizontally
AMD and Nvidia Ada GPU Specifications
Graphics CardRX 7900 XTXRX 7900 XTRX 6950 XTRTX 4090RTX 4080RTX 3090 TiRTX 3080 Ti
ArchitectureNavi 31Navi 31Navi 21AD102AD103GA102GA102
Process TechnologyTSMC N5 + N6TSMC N5 + N6TSMC N7TSMC 4NTSMC 4NSamsung 8NSamsung 8N
Transistors (Billion)45.6 + 6x 2.0545.6 + 5x 2.0526.876.345.928.328.3
Die size (mm^2)300 + 222300 + 185519608.4378.6628.4628.4
CUs / SMs968480128768480
GPU Shaders122881075251201638497281075210240
AI / Tensor Cores19216880512304336320
Ray Tracing Units968480128768480
Boost Clock (MHz)2500240023102520250518601665
VRAM Speed (Gbps)2020182122.42119
VRAM (GB)24201624162412
VRAM Bus Width384320256384256384384
L2 Cache9680128726466
ROPs192192128176112112112
TMUs384336320512304336320
TFLOPS FP3261.451.623.782.648.74034.1
TFLOPS FP16 (FP8/INT8)123 (123)103 (103)47.4661 (1321)390 (780)160 (320)136 (273)
Bandwidth (GBps)96080057610087171008912
TBP (watts)355315335450320450350
Launch DateDec-22Dec-22May-22Oct-22Nov-22Mar-22Jun-21
Launch Price$999 $899 $1,099 $1,599 $1,199 $1,999 $1,199

There's a lot to unpack in the specs, but we'll mostly focus on AMD's new chips. The RX 7900 XTX has the fully enabled Navi 31 GCD (Graphics Compute Die) along with six MCDs (Memory Cache Dies), while the 7900 XT disables a dozen compute units (CUs) in the GCD and one of the MCDs is fused off. Technically there are still six MCD chips present, to ensure even mounting pressure from the heatsink, but one of them is fused off (it could be a non-functional MCD).

The GPU shader counts are where things start to get a bit different from other architectures. AMD says there are still 64 Streaming Processors (SP) per CU, but there are now four SIMD32 vector units per CU as well — two of which can only process FP32 or Matrix operations and not INT32. We're going to call each of these a GPU shader, which goes along with AMD's peak throughput data of 61.4 teraflops FP32 on the 7900 XTX. This is similar to what Nvidia did with Ampere (and now Ada), so just know that the official SP count is no longer the same as the potential GPU shaders count.

We've also received some clarification on the "AI Accelerators" that are part of the RDNA 3 architecture. The short summary is that they repurpose the SIMD32 units to do matrix operations instead of FP32 (or FP16). They also support BF16 (16-bit brain-float) formats and INT8 alongside FP16. All three of those (FP16/BF16/INT8) have the same peak throughput that's double the FP32 single-precision floating-point throughput.

What's the difference between the previous half-precision FP16 shader support and the AI Accelerator FP16 support? Basically, it comes down to optimizing throughput and reducing power consumption, with some new instructions that are supported in matrix mode. Obviously, the peak FP16/BF16 rates are significantly lower than what the RTX 4080 and 4090 can deliver. Finding software that specifically uses the AI Accelerator on AMD's RDNA 2 / RDNA 3 GPUs is also proving difficult right now, so we may need to revisit the subject at a later date.

(Image credit: AMD)

Of course, one of the big deals with RDNA 3 and Navi 31 is the move to a chiplet architecture. Separating the memory and cache from the rest of the GPU functionality helps to bring prices down, or at least it reduces AMD's costs. It could also open the door to higher performance future designs. For now, however, it's important to note that all indications are that the use of chiplets does represent something of a performance compromise.

Not to jump ahead here, but if you look at the paper specs and see 160% more theoretical compute and 67% more memory bandwidth compared to the RX 6950 XT, clearly that doesn't really match up with AMD's own benchmarks that suggest more like a 50–80 percent improvement in individual games (around 60% on average). We feel some of that comes from additional overhead and latency associated with the use of chiplets. In other words, RDNA 3 chiplets feel a bit more like Zen 2 chiplets rather than, say, Zen 3 or Zen 4 chiplets where AMD's Ryzen CPUs really hit their stride.

There's another minor change from the previously published specifications on the RX 7900 XT. When first announced, AMD listed a 300W TBP (Total Board Power) rating, but it has since increased that to 315W. It says it found there was a decent increase in performance for a relatively minor bump in power use.

Finally, let's talk pricing and potential performance. We weren't particularly happy with Nvidia's "step down" RTX 4080 ushering in higher prices on the xx80 model GPUs, but AMD's RX 7900 XT is arguably just as bad. On paper, it has 17% less memory and memory bandwidth, and 16% less computational performance. That's fine on its own, but the price savings are only 10%. In other words, just like the RTX 4090 may actually be the better "deal" from Nvidia, we think most people eyeing the RX 7900 XT will be better off spending an extra $100 for the XTX model. 

Jarred Walton

Jarred Walton is a senior editor at Tom's Hardware focusing on everything GPU. He has been working as a tech journalist since 2004, writing for AnandTech, Maximum PC, and PC Gamer. From the first S3 Virge '3D decelerators' to today's GPUs, Jarred keeps up with all the latest graphics trends and is the one to ask about game performance.

  • -Fran-
    Thanks for the review!

    I'm reading it now, but I've watched numbers in other places. My initial reaction is lukewarm* TBH. I expected a bit more, but they're not terrible either. They did fall short of AMD's promise though. They indeed oversell the capabilities on raster, but were pretty on point for RT increases.

    Still, this card better have a "fine wine" effect down the line and the MSRP may just be well justified. This being said, it is still too expensive. for what it is.

    Regards.
    Reply
  • spongiemaster
    That's probably because this is the most competitive AMD has been in the consumer graphics market in quite some time.

    Not sure how this was determined, but I would argue this is a step backwards in almost every situation from the 6000 series. Also, it should be pointed out that there is something going on with the power consumption of the 7000 series in non-gaming situations that will affect many users. Looks like the memory isn't down clocking or something.

    Reply
  • JarredWaltonGPU
    spongiemaster said:
    Not sure how this was determined, but I would argue this is a step backwards in almost every situation from the 6000 series. Also, it should be pointed out that there is something going on with the power consumption of the 7000 series in non-gaming situations that will affect many users. Looks like the memory isn't down clocking or something.
    I don't do a ton of power testing scenarios, so I'd have to look into that more... and I really need to go sleep. As for the "being competitive," AMD is pretty much on par with Nvidia's best in rasterization (similar to 6000-series), and it's at least narrowed the gap in ray tracing. Or maybe that's just my perception? Anyway, since basically Pascal, it's felt like AMD GPUs have been very behind Nvidia. Nvidia offers more performance and more features, at an admittedly higher price.
    Reply
  • Colif
    Steve appears to have recorded a 750watt transient on the xtx which makes me sit back and wonder what PSU you need. though I am not sure if that is peak system power or just the gpu itself.
    Reply
  • Elusive Ruse
    A bit of a cynical Pros/Cons section, no mention of XTX being a much better value over 4080 which is its direct competition?
    Performance falls within expected margins (reasonable expectations, not that of crazed fanboys). Beating 4080 in rasterization and falling short in RT and professional uses. I don't quite care for RT but the performance gap in Blender e.g. is still eyepopping. I have heard of Blender 3.5 offering big improvements, yet that's not the current reality of things. I also doubt this will be a better story for RNDA 3 in Maya (Arnold) either.
    Reply
  • shADy81
    I assume the 750 W transient GN recorded is for the full system? TPU are showing 455W spike for the XTX and 412 for the XT, lower than 6900 and 6800 by quite some way on their charts.

    They also downgraded their PSU recommendation to 650 W for both cards. Was 1000W on the 6900. I think I'd feel a bit close on 650 W even allowing for a good quality unit being able to supply more than rated for short times. 650 would surely be way to close for a 13900K system, what do they know that I dont?
    Reply
  • zecoeco
    spongiemaster said:
    Not sure how this was determined, but I would argue this is a step backwards in almost every situation from the 6000 series. Also, it should be pointed out that there is something going on with the power consumption of the 7000 series in non-gaming situations that will affect many users. Looks like the memory isn't down clocking or something.


    This is actually a bug that was already reported to AMD and they're already working on a fix.
    Reply
  • zecoeco
    "Chiplets don't actually improve performance (and may hurt it)"
    How on earth is this even a con? who said chiplets are for performance? chiplets are for cost saving.
    But what did y'all expect ? You just can't complain for this price point.. there you go, chiplets saved you $200 bucks + gave you 24GB of VRAM as bonus (versus 16GB on 4080)
    It is meant for GAMING so don't expect productivity performance, for many reasons including nvidia's cuda cores that has every major software optimized for it.
    RDNA is going the right direction with chiplets.. in an industry of increasing costs year after year.
    Chiplet design is a solution, and not a new groundbreaking feature that's meant to boost performance.
    Sadly, instead of working on the problem, nvidia decided to give excuses such as "Moore's law is dead".
    Reply
  • salgado18
    But for under a grand, right now the RX 7900 XTX delivers plenty to like and at least keeps pace with the more expensive RTX 4080. All you have to do is lose a good sized chunk of ray tracing performance, and hope that FSR2 can continue catching up to DLSS.
    That, I believe, is the reason AMD won't increase too much their market share in this generation. Yes, rasterization is comparable, so are power, memory, price and even upscaling performance/quality. But it is a bad card for raytracing, or at least that's the message, and between a full card and a crippled card, people will prefer the fully featured one. I know designing GPUs is a monstrously complex task, but they really needed to up their RT performance by at least 3x to be competitive. Now they will keep being "bang-for-buck", which is nice, but never "the best".

    Edit: by some rough calcs, if the XTX is ~40% faster without RT than the 6950, and ~50% faster with RT, then the generational improvement is ~7%? If so, then that's hardly any improvement at all. Great cards and all that, but I'm very disappointed with the lack of focus on RT.
    Reply
  • btmedic04
    salgado18 said:
    That, I believe, is the reason AMD won't increase too much their market share in this generation. Yes, rasterization is comparable, so are power, memory, price and even upscaling performance/quality. But it is a bad card for raytracing, or at least that's the message, and between a full card and a crippled card, people will prefer the fully featured one. I know designing GPUs is a monstrously complex task, but they really needed to up their RT performance by at least 3x to be competitive. Now they will keep being "bang-for-buck", which is nice, but never "the best".

    Edit: by some rough calcs, if the XTX is ~40% faster without RT than the 6950, and ~50% faster with RT, then the generational improvement is ~7%? If so, then that's hardly any improvement at all. Great cards and all that, but I'm very disappointed with the lack of focus on RT.

    AMD has a definite physical size advantage though which is something thats applicable to quite a few folks (myself included.) I hear what you are saying, but as a 3090 owner, 3090-like RT performance from the 7900xtx is still quite good for most people.
    Reply