• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

ASRock Radeon RX 7900 XTX Taichi

I have a general methodology question, isn't Doom Eternal skewing the results of the review because it has a frame rate so much higher than any other game, and therefore the "average framerate" data is almost meaningless because it is being "pulled up" by Doom Eternal so much? i mean, 1 FPS of Doom Eternal does not equal 1 FPS of CP2077...the first is much "easier" to produce, hence the observed higher frame rate. i think the outliers like Doom Eternal need to be removed from the suite, or at least not included in the average.
in my opinion, the problem is exacerbated even more when the "relative performance" table is derived from the same "average framerate" data. making them both seem higher than they actually are in real life situations.


If I declare DOOM an outlier, then the game with the lowest FPS is an outlier too.

and here I have removed everything that could be considered an outlier, and I could remove even more.

Perhaps stating the removal of the 2 titles with the lowest FPS and the 2 titles with the highest FPS from the averages, EVEN BEFORE displaying the benchmarks results: the idea is to remove outliers, in either direction.

This, for reviews with A LOT of game titles: for reviews with few game titles, suggest reducing the game titles removal to 1 in either direction.

My 2 cents.
 
FYI there are dozens of 7900 xtx in stock at $999 price point currently.
 
So it's still much behind NVIDIA when comes to RT.
Games will mostly have RT in the future.
Hopefully AMD can improve this.
 
RDNA3 is a MASSIVE leap in RT performance for AMD. It beats the 3090 Ti.

The XTX is the most in demand card of the new generation. Every card always sold out, even way above MSRP. Highest perf/dollar.

It has been more than 2 months, would like to see more on the shelf. An actual one on AMD's website once would be nice.
RDNA3 didn't improve actual RT performance by any meaningful amount. It is actually the most disappointing performance improvement for RDNA3 IMO. The 3090 Ti is actually much more efficient at raytracing, losing ~45% baseline performance, compared to the 7900 XTX, losing ~55%.

Best case is a few percentage(3-5%) improved efficiency over base non-RT performance compared to that of RDNA2. Nvidia's 4xxx series also was a disappointing improvement in RT efficiency, only they started out at a much better baseline efficency than that of AMD.

I was expecting a larger leap in RT efficiency with RDNA3 but it didn't happen. Hoping RDNA4 will increase efficiency further to get it more in line with where Nvidia is at.
 
So it's still much behind NVIDIA when comes to RT.
Games will mostly have RT in the future.
Hopefully AMD can improve this.

That catch is in the future. We are legit not going to see ray tracing or even the full benefits from X3D cache in the broad sense till the PS6 and others fully have them.
 
While I know this is for microcenter hopefully this is a foreshadowing of things to come everywhere.
I'll keep you updated for more.

close to msrp
at $1049

 
Last edited:
So it's still much behind NVIDIA when comes to RT.
Games will mostly have RT in the future.
Hopefully AMD can improve this.
So far there is no definitive trend that RT will go anywhere. Right now it is way too early to tell where rendering will end up. By the way there is another trend that is in total opposition to RT: high refresh rate esports gaming. You will not see full or even partial path RT over 100 fps any time soon.

In addition ALL RT implementations today are a gimmick at best. I can see the vast majority enabling RT at first and then disabling it when the massive performance hit ruins the gaming experience versus the barely noticeable graphics quality improvement.
 
LED OFF Switch? All other brands SHOULD TAKE NOTE.
RGB should be factory disabled really.

Those that want christmas lighting can install software, but don't force down peoples throat who HATE it.
 
While I know this is for microcenter hopefully this is a foreshadowing of things to come everywhere.
aha .. good info, in-store pickup only though, but your point is taken. If I can get ref for 999 and ASRock for 1120 I'd buy ref unless I absolutely want the low noise
 
Great card but reaching the 4080 price point. I believe the Red Devil or Merc 310 at the $1,049 price point are better value options.
 
Last edited:
No doubt about it this is a good card... Thanks for the review W1z.

This is the only generation I've purchased a reference card. AMD met their pricing on the reference card, and it really made the investment worth while versus a 4080, or for that matter a 4090.
 
RDNA3 didn't improve actual RT performance by any meaningful amount. It is actually the most disappointing performance improvement for RDNA3 IMO. The 3090 Ti is actually much more efficient at raytracing, losing ~45% baseline performance, compared to the 7900 XTX, losing ~55%.

Best case is a few percentage(3-5%) improved efficiency over base non-RT performance compared to that of RDNA2. Nvidia's 4xxx series also was a disappointing improvement in RT efficiency, only they started out at a much better baseline efficency than that of AMD.

I was expecting a larger leap in RT efficiency with RDNA3 but it didn't happen. Hoping RDNA4 will increase efficiency further to get it more in line with where Nvidia is at.
that's completely untrue, go read the reviews on this site again
 
Last edited:
"This is my fifth review of a Radeon RX 7900 XT/XTX Navi 31 card. Previously I've tested the ASUS RX 7900 XTX TUF, XFX RX 7900 XTX Merc 310, AMD RX 7900 XTX and AMD RX 7900 XT."

@W1zzard, did't you also review the Sapphire Nitro+, which makes the TAICHI your sixth, not fifth, 7000 series card? And isn't the Sapphire your best OC performer, not the Taichi?
I claim I can review these products but can't even count to six t.t .. fixed ;)

Yeah Sapphire OC'd best. Actually in the retest with 3DMark TimeSpy Extreme (which I'm using to replace Unigine the XFX card did a tiny bit better.
 
that's completely untrue, go read the reviews on this site again
https://tpucdn.com/review/asrock-radeon-rx-7900-xtx-taichi/images/control-rt-3840-2160.png
Difference from non-raytracing FPS:
7900XTX: -55%
6900XT: -55%
3090 Ti: -41%
https://tpucdn.com/review/asrock-radeon-rx-7900-xtx-taichi/images/cyberpunk-2077-rt-3840-2160.png
Difference from non-raytracing FPS:
7900XTX: -67%
6900XT: -70%
3090 Ti: -50%

Are you just considering 'better raytracing performance' as the relative FPS at the top of the raytracing page? If that's the case, basically every single generation is going to have 'better raytracing performance' than the generation before it.

The real measurement of better raytracing performance is the base FPS without raytracing compared to the FPS with raytracing, and using that metric, RDNA3 had an embarrassingly small improvement considering they specifically targeted raytracing when updating RDNA3 architecture.
 
I just installed this card for my new build this weekend and am excited to take it through its paces. Unfortunately the game I want to play most right now (Hogwarts Legacy) is not getting on nicely with the version of Windows I'm on, but that's not the card's fault. There are also some strange issues in the game with AMD cards with foliage and strange outlines around the characters. Neither of these are necessarily AMD's fault, but still slightly disappointing to have to wait a while to play the only game I wish to play at the moment.
 
https://tpucdn.com/review/asrock-radeon-rx-7900-xtx-taichi/images/control-rt-3840-2160.png
Difference from non-raytracing FPS:
7900XTX: -55%
6900XT: -55%
3090 Ti: -41%
https://tpucdn.com/review/asrock-radeon-rx-7900-xtx-taichi/images/cyberpunk-2077-rt-3840-2160.png
Difference from non-raytracing FPS:
7900XTX: -67%
6900XT: -70%
3090 Ti: -50%

Are you just considering 'better raytracing performance' as the relative FPS at the top of the raytracing page? If that's the case, basically every single generation is going to have 'better raytracing performance' than the generation before it.

The real measurement of better raytracing performance is the base FPS without raytracing compared to the FPS with raytracing, and using that metric, RDNA3 had an embarrassingly small improvement considering they specifically targeted raytracing when updating RDNA3 architecture.
From the same charts you linked, RTX 30 to 40 doesn't evolve much as well. It's what? 3% better at most?
 
RDNA3 is a MASSIVE leap in RT performance for AMD. It beats the 3090 Ti.

The XTX is the most in demand card of the new generation. Every card always sold out, even way above MSRP. Highest perf/dollar.

It has been more than 2 months, would like to see more on the shelf. An actual one on AMD's website once would be nice.
This card's stock seems to ebb and flow. I bought mine off of Newegg at MSRP with no back-ordering or waiting whatsoever. Note that the week before it was back-ordered, however.

https://tpucdn.com/review/asrock-radeon-rx-7900-xtx-taichi/images/control-rt-3840-2160.png
Difference from non-raytracing FPS:
7900XTX: -55%
6900XT: -55%
3090 Ti: -41%
https://tpucdn.com/review/asrock-radeon-rx-7900-xtx-taichi/images/cyberpunk-2077-rt-3840-2160.png
Difference from non-raytracing FPS:
7900XTX: -67%
6900XT: -70%
3090 Ti: -50%

Are you just considering 'better raytracing performance' as the relative FPS at the top of the raytracing page? If that's the case, basically every single generation is going to have 'better raytracing performance' than the generation before it.

The real measurement of better raytracing performance is the base FPS without raytracing compared to the FPS with raytracing, and using that metric, RDNA3 had an embarrassingly small improvement considering they specifically targeted raytracing when updating RDNA3 architecture.
It's really quite simple- AMD intentionally did not make Ray Tracing their first priority for this generation, and if it's important to you, don't buy this card. I understood that when I bought mine last week, and accept that Ray Tracing will be 10-15 percent below the 4080. I'm okay with that.
 
Last edited:
From the same charts you linked, RTX 30 to 40 doesn't evolve much as well. It's what? 3% better at most?
The 40 series was also a disappointment in terms of RTX improvement. The saving grace in terms of raytracing performance is they were in a much better place than AMD, and because of the marginal improvements of both the AMD & Nvidia series this generation in Raytracing, they're still ahead by the same amount.
It's really quite simple- AMD intentionally did not make Ray Tracing their first priority for this generation, and if it's important to you, don't buy this card. I understood that when I bought mine last week, and accept that Ray Tracing will be 10-15 percent below the 4080. I'm okay with that.
My reply was to someone saying RDNA3 made significant gains in raytracing. I was just pointing out that it made no significant gains over RDNA2 in terms of actual raytracing. The only reason there is the perception of significant raytracing gains is because of the significant rasterization gains.

It was pretty clear AMD was attempting to reconfigure RDNA3 to improve raytracing. Unfortunately that work didn't translate into a real-world improvements to raytracing performance. Here's a quote from AMD's RDNA3 press release:
  • Dedicated AI Acceleration and Second-Generation Raytracing – New AI instructions and increased AI throughput deliver up to 2.7X more performance than the previous AMD RDNA 2 architecture, while second-generation raytracing technology delivers up to 1.8X more performance than the previous generation.
That press release was a bit misleading though. As they said 7900XTX would be 1.7x faster than the 6950XT, so I guess they meant to imply that it had just a 10% improvement over RDNA2 in raytracing performance, but that 10% improvement rarely translates into a 10% improvement in games. Typically we're seeing a 0-5% improvement over RDNA2.
 
There are also some strange issues in the game with AMD cards with foliage
Are you on the latest driver? Supposedly that was fixed with 23.2.1.
 
This card's stock seems to ebb and flow. I bought mine off of Newegg at MSRP with no back-ordering or waiting whatsoever. Note that the week before it was back-ordered, however.


It's really quite simple- AMD intentionally did not make Ray Tracing their first priority for this generation, and if it's important to you, don't buy this card. I understood that when I bought mine last week, and accept that Ray Tracing will be 10-15 percent below the 4080. I'm okay with that.
In Canada you just can't buy the 7900 XTX. 2 months later. Very annoying. I'm not paying $1500 for the $1300 CAD card.
 
Are you on the latest driver? Supposedly that was fixed with 23.2.1.
I'm on 23.2.2. The problem isn't the foliage anymore; it's with an outline around the characters and the mouths not moving. Several sources indicate that rolling back Windows should help this (which I'm unwilling to do). I've also confirmed the fix that if you turn on RT shadows to Ultra it at least fixes the weird outline around the characters, but not the mouths not moving.

In Canada you just can't buy the 7900 XTX. 2 months later. Very annoying. I'm not paying $1500 for the $1300 CAD card.
Your best bet might be to try to keep trying the AMD site. Try early in the mornings, particularly towards the end of the work week.
 
Gpu prices are falling fyi
879.99
$
829
.99
Sale Ends in 14 hours

Lowest Price
in 30 days
Save: $50.00 (5%)
30 Days Hassle-Free Returns
ASRock Phantom Gaming Radeon RX 7900 XT 20GB GDDR6 PCI Express 4.0 ATX Video Card RX7900XT PG 20GO
 
Back
Top