AMD RX 7700 XT GPU Review & Benchmarks vs. 7800 XT, 6800 XT, RTX 4060 Ti, & More
video description
Date: 2023-09-07
Related videos
Comments and reviews: 20
TheSupremeSkill
I would choose the RX 7800 XT over the RX 6800 XT, because the calculation of collisions for ray tracing in RDNA3 is streamlined, so I expect in some future titles - perhaps this is reality already in some current ones - the advantage might show, no matter that it has less CUs. Also the power consumption is in the ballpark of acceptable, rather than the 300 W, which is for many not optimally ventilated cases a bit much, and a bit much in general for a computer part in my opinion. I think it s not too wise to invest in the old gen in this case when the newer one offers a few advantages with such a small performance difference, but that s just my take.
Both cards will serve nicely, even at 4K with upscaling, but alas, for the price you pay for the current gen cards, you might surprise how fast they start to feel slow compared to previous purchases in earlier gens, because after all, this is the performance that was put on the table three years ago. The console generation doesn t shift any time soon, but refreshed models with RDNA3 might arrive soon enough, which may up the GPU requirements once more for the best experience. :O No worry though, eventually some generation will be more affordable and who knows what the Battlemage brings up.
As for the RX 7700 XT, the upsell is obvious, so the model awaits a price cut. Totally not a bad GPU being so close to the RTX 4070. It will likely sell just fine down the track.
reply
I would choose the RX 7800 XT over the RX 6800 XT, because the calculation of collisions for ray tracing in RDNA3 is streamlined, so I expect in some future titles - perhaps this is reality already in some current ones - the advantage might show, no matter that it has less CUs. Also the power consumption is in the ballpark of acceptable, rather than the 300 W, which is for many not optimally ventilated cases a bit much, and a bit much in general for a computer part in my opinion. I think it s not too wise to invest in the old gen in this case when the newer one offers a few advantages with such a small performance difference, but that s just my take.
Both cards will serve nicely, even at 4K with upscaling, but alas, for the price you pay for the current gen cards, you might surprise how fast they start to feel slow compared to previous purchases in earlier gens, because after all, this is the performance that was put on the table three years ago. The console generation doesn t shift any time soon, but refreshed models with RDNA3 might arrive soon enough, which may up the GPU requirements once more for the best experience. :O No worry though, eventually some generation will be more affordable and who knows what the Battlemage brings up.
As for the RX 7700 XT, the upsell is obvious, so the model awaits a price cut. Totally not a bad GPU being so close to the RTX 4070. It will likely sell just fine down the track.
reply
Slick
Keep in mind that the vast majority of other reviewers (such as HUB and Daniel Owens), who tested more modern demanding games, found the 7800xt consistently ahead of the 6800xt overall, by roughly 2-3% on average in Raster, and 5-6% in RT. Not a huge margin to be sure, but when coupled with it being strictly better in every other way (Power Efficiency, Av1 Encoding, Memory Bandwidth advantage, better features with Hyper-Rx Fluid Frames and Anti-Lag+), including the Price on average (some sales at the 500 mark can be found for the 6800xt now, but on average it has been at the 510-530 mark these past few months). If the 6800xt was 50 cheaper than the 7800xt, it would be a strong alternative assuming the stock isn't fully dried up (and it all but surely will be in the coming weeks). In reality though if the 7800xt and 6800xt are at a similar price, it would be foolish to buy the 6800xt over it, despite allegedly being Better according to the usually solid GN benchmarks. Remember just look at the averages of all the reliable benchmarks and reviewers, and its clear that the 7800xt is Strictly Better than the 6800xt, despite this outlier.
reply
Keep in mind that the vast majority of other reviewers (such as HUB and Daniel Owens), who tested more modern demanding games, found the 7800xt consistently ahead of the 6800xt overall, by roughly 2-3% on average in Raster, and 5-6% in RT. Not a huge margin to be sure, but when coupled with it being strictly better in every other way (Power Efficiency, Av1 Encoding, Memory Bandwidth advantage, better features with Hyper-Rx Fluid Frames and Anti-Lag+), including the Price on average (some sales at the 500 mark can be found for the 6800xt now, but on average it has been at the 510-530 mark these past few months). If the 6800xt was 50 cheaper than the 7800xt, it would be a strong alternative assuming the stock isn't fully dried up (and it all but surely will be in the coming weeks). In reality though if the 7800xt and 6800xt are at a similar price, it would be foolish to buy the 6800xt over it, despite allegedly being Better according to the usually solid GN benchmarks. Remember just look at the averages of all the reliable benchmarks and reviewers, and its clear that the 7800xt is Strictly Better than the 6800xt, despite this outlier.
reply
Andrew
I feel like I'm not the only one that feels that the current gen GPUs from both Nvidia and AMD are mostly underwhelming given theIR pricing. If anything the current gen GPUs continue to prove that that the last gen GPUs are still extremely relevant and are often much better value for your money even if they don't come with the newest features like DLSS 3.0 or a newer architecture.
There used to be a time where I'd consider buying new GPUs, but right now given how you could get a used last gen GPU that many times has MUCH better rasterization performance than a current gen GPU for around the same price to me is a no brainer.
I hope that GN continues to talk about how new GPUs compare to what you could buy in the used market because it's a very important component for a lot of folks for considering the value aspect. Yes in as early as 6 months to a year the used GPU market will be different, but I'd argue its still worth addressing in GPU reviews because many of us who don't have a burning hole in our pocket to buy an RTX 4090 at the 1599 MSRP are looking at used GPUs so we can get the most bang for our buck.
reply
I feel like I'm not the only one that feels that the current gen GPUs from both Nvidia and AMD are mostly underwhelming given theIR pricing. If anything the current gen GPUs continue to prove that that the last gen GPUs are still extremely relevant and are often much better value for your money even if they don't come with the newest features like DLSS 3.0 or a newer architecture.
There used to be a time where I'd consider buying new GPUs, but right now given how you could get a used last gen GPU that many times has MUCH better rasterization performance than a current gen GPU for around the same price to me is a no brainer.
I hope that GN continues to talk about how new GPUs compare to what you could buy in the used market because it's a very important component for a lot of folks for considering the value aspect. Yes in as early as 6 months to a year the used GPU market will be different, but I'd argue its still worth addressing in GPU reviews because many of us who don't have a burning hole in our pocket to buy an RTX 4090 at the 1599 MSRP are looking at used GPUs so we can get the most bang for our buck.
reply
Aaron
So, Steve, this is unrelated to the review itself and I am questioning your comments about percent math being odd . You said it in both videos, and I'm not following why you've said it. Percent math is quite normal so long as you are consistent with the comparison. I actually fear the way you're doing it is the opposite and makes it more confusing.
So, % difference = (State_2 - State_1)/State_1. If GPU 1, the thing we're trying to compare, so in this case the 7700 XT, got 100 FPS. If you say 10% higher (+10%), that means 10 FPS more (110 FPS). If you say 10% lower (-10%), it's still -10 FPS (90 FPS). The math is actually very consistent. Now, if you change your base comparison to now use a different denominator, that's when things get screwy. If you now did GPU 2 (110 FPS) as your basis (which is what you've done in the charts by flipping the comparisons), the math is (100-110)/110 -9%.
Perhaps I'm missing your point, but I'd love to understand where you're coming from. Maybe I'm not thinking about this in the same way you are.
reply
So, Steve, this is unrelated to the review itself and I am questioning your comments about percent math being odd . You said it in both videos, and I'm not following why you've said it. Percent math is quite normal so long as you are consistent with the comparison. I actually fear the way you're doing it is the opposite and makes it more confusing.
So, % difference = (State_2 - State_1)/State_1. If GPU 1, the thing we're trying to compare, so in this case the 7700 XT, got 100 FPS. If you say 10% higher (+10%), that means 10 FPS more (110 FPS). If you say 10% lower (-10%), it's still -10 FPS (90 FPS). The math is actually very consistent. Now, if you change your base comparison to now use a different denominator, that's when things get screwy. If you now did GPU 2 (110 FPS) as your basis (which is what you've done in the charts by flipping the comparisons), the math is (100-110)/110 -9%.
Perhaps I'm missing your point, but I'd love to understand where you're coming from. Maybe I'm not thinking about this in the same way you are.
reply
WarcraftTechGuy
Just something to note that I realized an issue with some of your graphs that makes viewing this data quite difficult. Not sure if you do this for a purpose or if it's because your all tired or even if this might be something you've always done but I just now was made aware of it...
Some of the FPS graphs you shared have their upper limits set too high to visualize the comparisons. They range from 100% to over 200% that of the highest achived performance in the graph, providing ALOT of dead space and DENSE bar graphs.
For example, Dying Light 2's 4k graph has its upper limit is set to 200FPS, 2.6 times higher than the best performing card at 76FPS. Trying to see the 11% improvement from 37.7FPS to 41.7FPS is VERY hard when the scale is set so high that an 11% improvement is 1/50th of the entire bar graph...
Games Graphs that showcase this:
Dying Light 2 All Resolutions
Resident Evil 4k & 1440p
Total War Warhamer 4k & 1440p
Tomb Raider 4k
Horizon Zero Dawn 4k
FFXIV 4k
All Ray Tracing graphs
reply
Just something to note that I realized an issue with some of your graphs that makes viewing this data quite difficult. Not sure if you do this for a purpose or if it's because your all tired or even if this might be something you've always done but I just now was made aware of it...
Some of the FPS graphs you shared have their upper limits set too high to visualize the comparisons. They range from 100% to over 200% that of the highest achived performance in the graph, providing ALOT of dead space and DENSE bar graphs.
For example, Dying Light 2's 4k graph has its upper limit is set to 200FPS, 2.6 times higher than the best performing card at 76FPS. Trying to see the 11% improvement from 37.7FPS to 41.7FPS is VERY hard when the scale is set so high that an 11% improvement is 1/50th of the entire bar graph...
Games Graphs that showcase this:
Dying Light 2 All Resolutions
Resident Evil 4k & 1440p
Total War Warhamer 4k & 1440p
Tomb Raider 4k
Horizon Zero Dawn 4k
FFXIV 4k
All Ray Tracing graphs
reply
Freelance
While yes, when you look at other benchmarking videos of the 4060 Ti you know it's a bad value even compared to the 6800XT, this video doesn't really show that as well as it could because you didn't test against the 16GB variant as far as I saw. On a lot of older games and/or at lower resolutions, sure that won't matter, but at higher resolutions and/or in newer games the 16GB VRAM is going to make a difference, especially in your 1% and 0.1% lows mostly because there will be less loading and unloading data. I'm seeing a lot of newer games use 9GB or more of VRAM.
I looked at some separate benchmarks and saw the differences for myself and most seemed to show the 4060Ti 16GB variant being absolutely destroyed by the 6800 XT (which makes me feel good as the owner of a 6800 XT) but I was kind of disappointed that you didn't give the 4060Ti the best possible chance so as to not even appear to cheapen your stance about its value.
reply
While yes, when you look at other benchmarking videos of the 4060 Ti you know it's a bad value even compared to the 6800XT, this video doesn't really show that as well as it could because you didn't test against the 16GB variant as far as I saw. On a lot of older games and/or at lower resolutions, sure that won't matter, but at higher resolutions and/or in newer games the 16GB VRAM is going to make a difference, especially in your 1% and 0.1% lows mostly because there will be less loading and unloading data. I'm seeing a lot of newer games use 9GB or more of VRAM.
I looked at some separate benchmarks and saw the differences for myself and most seemed to show the 4060Ti 16GB variant being absolutely destroyed by the 6800 XT (which makes me feel good as the owner of a 6800 XT) but I was kind of disappointed that you didn't give the 4060Ti the best possible chance so as to not even appear to cheapen your stance about its value.
reply
ProgressivistMeans
It's at this point I believe that AMD and NVIDIA have an understanding with pricing nowadays. AMD doesn't undercut too much, NVIDIA keeps their prices at MSRP. Both Win . I mean, what else can it be? AMD's silicon cost in comparison to NVIDIA is much cheaper. They choices they made on where to cut down dies, their naming choices, and unwillingness to go for the throat has me in wonder. They've both decided they have been undercharging and so both have raised prices significantly and take what they can get from gamers.
Shrug I bought a 2070 against my better judgement after waiting 3 generations only to get burned badly. I'm waiting another gen, I'm going to assume this is the phoned in generation and next generation will have better performance. I'll play on medium settings, thanks. Go intel!
reply
It's at this point I believe that AMD and NVIDIA have an understanding with pricing nowadays. AMD doesn't undercut too much, NVIDIA keeps their prices at MSRP. Both Win . I mean, what else can it be? AMD's silicon cost in comparison to NVIDIA is much cheaper. They choices they made on where to cut down dies, their naming choices, and unwillingness to go for the throat has me in wonder. They've both decided they have been undercharging and so both have raised prices significantly and take what they can get from gamers.
Shrug I bought a 2070 against my better judgement after waiting 3 generations only to get burned badly. I'm waiting another gen, I'm going to assume this is the phoned in generation and next generation will have better performance. I'll play on medium settings, thanks. Go intel!
reply
Brandon
they literally are taking the same approach as nvidia and relying on software locked in behind generations. i have had issues with stuttering and driver crashes with my 6700xt if i try to do anything not factory with it. even locking the clock within 100mhz in a downclocked state causes crashes. the hyper-rx software suite is locked behind the new generation of cards and it legitimately is awful because that is the one thing i can say i cannot stand about the 6000 series is the god awful frame pacing and input lag. they basically said if u want high quality fsr 3 that doesnt give u 120ms input lag u must buy into the new generation that performs at best 5% better to 15% worse.
reply
they literally are taking the same approach as nvidia and relying on software locked in behind generations. i have had issues with stuttering and driver crashes with my 6700xt if i try to do anything not factory with it. even locking the clock within 100mhz in a downclocked state causes crashes. the hyper-rx software suite is locked behind the new generation of cards and it legitimately is awful because that is the one thing i can say i cannot stand about the 6000 series is the god awful frame pacing and input lag. they basically said if u want high quality fsr 3 that doesnt give u 120ms input lag u must buy into the new generation that performs at best 5% better to 15% worse.
reply
Stephen
AWESOME - Thanks to a history of banging reviews like this I totally sidestepped the GPU pricing hell and 18 months ago upgraded from a 4690K/ GTX 1070 to a 3600/ 6600XT. First time in decades I'd stepped from the Intel/ NVidia slow lane of innovation and gouging.
Nine months later I bought my first ever 2nd hand video card. A Red Dragon 6800 for 382 ( 650 NZD) which after using MorePowerTools, undervolting and overclocking gets my boost clock to a stable 2460 in game !!!
So I'm looking at your 6800 XT results and going yeah baby'.
And I thought the 1070 was a fabulous bang for buck as it lasted me 4 years.
Cheers and thanks to you and the team, carry on.
reply
AWESOME - Thanks to a history of banging reviews like this I totally sidestepped the GPU pricing hell and 18 months ago upgraded from a 4690K/ GTX 1070 to a 3600/ 6600XT. First time in decades I'd stepped from the Intel/ NVidia slow lane of innovation and gouging.
Nine months later I bought my first ever 2nd hand video card. A Red Dragon 6800 for 382 ( 650 NZD) which after using MorePowerTools, undervolting and overclocking gets my boost clock to a stable 2460 in game !!!
So I'm looking at your 6800 XT results and going yeah baby'.
And I thought the 1070 was a fabulous bang for buck as it lasted me 4 years.
Cheers and thanks to you and the team, carry on.
reply
BReal-10EC
All of the last gen AMD GPUs at Newegg (I just checked) give me a 5 per customer purchase limit. That could be a sign they are close to running out of inventory. NVidia cards are still the normal 20 per customer limit.. So, is AMD playing the bottom end here? Affordable GPUs are needed bad, so I can see them bumping the margin on the 7700 XT thinking it will be in more demand than the 7800 XT. But it could also be due to yield. If the 7700 XT is the same die but just needs to be kneecapped to save a bad chip.. maybe their yield is too good on the 7800 XT side so they know the 7700 XT supply will be more limited?
reply
All of the last gen AMD GPUs at Newegg (I just checked) give me a 5 per customer purchase limit. That could be a sign they are close to running out of inventory. NVidia cards are still the normal 20 per customer limit.. So, is AMD playing the bottom end here? Affordable GPUs are needed bad, so I can see them bumping the margin on the 7700 XT thinking it will be in more demand than the 7800 XT. But it could also be due to yield. If the 7700 XT is the same die but just needs to be kneecapped to save a bad chip.. maybe their yield is too good on the 7800 XT side so they know the 7700 XT supply will be more limited?
reply
RandoCrypto
Steve and GN team, honest question. Just trying to understand, not throw shade. When testing GPUs, sometimes you re running into a CPU constrained scenario. This is mentioned by the presenter and obvious on the charts. Because this is the case and makes GPU reviews less accurate in showing the true differences between cards, why has a transition not been made to a faster CPU like a 13900K or 7800X3D? This would make a statement such as X GPU is X percent faster than X GPU more accurate because that percentage wouldn t be as influenced by other components in the system such as the CPU.
reply
Steve and GN team, honest question. Just trying to understand, not throw shade. When testing GPUs, sometimes you re running into a CPU constrained scenario. This is mentioned by the presenter and obvious on the charts. Because this is the case and makes GPU reviews less accurate in showing the true differences between cards, why has a transition not been made to a faster CPU like a 13900K or 7800X3D? This would make a statement such as X GPU is X percent faster than X GPU more accurate because that percentage wouldn t be as influenced by other components in the system such as the CPU.
reply
Leafhopper
I got a 6700XT in March and thought the card was good but, weird issue with AMD drivers made me get a RTX 4070 2 months ago. The weird driver issue I had would be sometimes it would crash and fully uninstall itself. Baffling but that probably is just a me problem. The card was very good, still have it but in it's box. Would have kept using it if not for that but, went back to Nvidia since the DPC issues it's had isn't as bad as I was having on my 2070 Super.
The 7800XT seems like a good card too and if you don't have bizarre driver issues like me it's probably really good.
reply
I got a 6700XT in March and thought the card was good but, weird issue with AMD drivers made me get a RTX 4070 2 months ago. The weird driver issue I had would be sometimes it would crash and fully uninstall itself. Baffling but that probably is just a me problem. The card was very good, still have it but in it's box. Would have kept using it if not for that but, went back to Nvidia since the DPC issues it's had isn't as bad as I was having on my 2070 Super.
The 7800XT seems like a good card too and if you don't have bizarre driver issues like me it's probably really good.
reply
koray
You guys still forgot, some countrys electry bill, like mine is extremely expensive. So when i choose a card i must consider power consumption too. Like US ppl don't care about electricity as it is almost free and it is always skipped in comparisons. Also, it is wrong to compare cards with raw power. Comparisons are still made with the old ways. Now it is necessary to compare the cards with the technologies they have. The arguments that nobody uses DLSS and Ray Tracing are also completely ridiculous. And lastly, the 4060 TI rip again writing is completely ridiculous.
reply
You guys still forgot, some countrys electry bill, like mine is extremely expensive. So when i choose a card i must consider power consumption too. Like US ppl don't care about electricity as it is almost free and it is always skipped in comparisons. Also, it is wrong to compare cards with raw power. Comparisons are still made with the old ways. Now it is necessary to compare the cards with the technologies they have. The arguments that nobody uses DLSS and Ray Tracing are also completely ridiculous. And lastly, the 4060 TI rip again writing is completely ridiculous.
reply
John
I still am shocked to see the A770 and A750 doing so well compared to the RX 7600. In some games, it pulls ahead by 10% or more, other titles it matches or lags behind. The only reason I didn't commit to one was because I have an older X370 motherboard I'd pair with it, which may or may not have ReBAR/SAM support available even after I update the BIOS. That, and I grabbed a 7600 for about 60 cheaper at Microcenter. Had they been the same price, I would have come back with an Intel GPU and tried that out for a while.
reply
I still am shocked to see the A770 and A750 doing so well compared to the RX 7600. In some games, it pulls ahead by 10% or more, other titles it matches or lags behind. The only reason I didn't commit to one was because I have an older X370 motherboard I'd pair with it, which may or may not have ReBAR/SAM support available even after I update the BIOS. That, and I grabbed a 7600 for about 60 cheaper at Microcenter. Had they been the same price, I would have come back with an Intel GPU and tried that out for a while.
reply
X2yt
While these cards don't perform as one would expect from new gen cards, there is an upside to it. Since the MSRP of these cards is lower than equivalent cards of old gen, but performance is the similar, when AMD inevitably drops the MSRP in 1-2 months like they usually do these days, these cards will be very good purchased for mid-range systems. Sure, the performace won't change much, maybe few percentage points here and there, but because of MSRP dropsl, the price to performance proposition will improve over time.
reply
While these cards don't perform as one would expect from new gen cards, there is an upside to it. Since the MSRP of these cards is lower than equivalent cards of old gen, but performance is the similar, when AMD inevitably drops the MSRP in 1-2 months like they usually do these days, these cards will be very good purchased for mid-range systems. Sure, the performace won't change much, maybe few percentage points here and there, but because of MSRP dropsl, the price to performance proposition will improve over time.
reply
rdspam
Reviewers:
The 7800XT is a terrible product because it s the same number as the 6800 at the same price (ignoring in was 150 more at launch). The 7700XT is so much better than the 6700, with the same number, so it s a great value.
Ignore the fact the evaluating the cards for what they are is the opposite.
Maybe 3D print some lower number stickers to make everything happy with the product. Slap a 7600 sticker on the 7800XT. Huge generational uplift, what an incredible product!!! , right?
reply
Reviewers:
The 7800XT is a terrible product because it s the same number as the 6800 at the same price (ignoring in was 150 more at launch). The 7700XT is so much better than the 6700, with the same number, so it s a great value.
Ignore the fact the evaluating the cards for what they are is the opposite.
Maybe 3D print some lower number stickers to make everything happy with the product. Slap a 7600 sticker on the 7800XT. Huge generational uplift, what an incredible product!!! , right?
reply
Besher
I figured it out! amd marketing department is either out of touch or genius
if it is intentional. they priced their cards in a way to increase sales of the older over supply of cards while still being highly competitive with nvidia, which might also be why they priced the two cards so closely to spike interest, so the first card you would want to buy would be the 6000 then the 7000 and lastly nvidia.
though them naming the 7800 a 7800xt should not be forgiven even if nvidia did it first
reply
I figured it out! amd marketing department is either out of touch or genius
if it is intentional. they priced their cards in a way to increase sales of the older over supply of cards while still being highly competitive with nvidia, which might also be why they priced the two cards so closely to spike interest, so the first card you would want to buy would be the 6000 then the 7000 and lastly nvidia.
though them naming the 7800 a 7800xt should not be forgiven even if nvidia did it first
reply
RavenFyre
It is a good card by today's standards and when competing with Nvidia's dumb pricing. Unfortunately due to GPU stagnation we should be much further along in general. The manufacturers are just cruising now. If AMD is to compete, they need a higher quality FSR upscaling when compared to DLSS since now we are dependent on these technologies to maintain framerate. I bet the 7850xt will be great.
reply
It is a good card by today's standards and when competing with Nvidia's dumb pricing. Unfortunately due to GPU stagnation we should be much further along in general. The manufacturers are just cruising now. If AMD is to compete, they need a higher quality FSR upscaling when compared to DLSS since now we are dependent on these technologies to maintain framerate. I bet the 7850xt will be great.
reply
Mattzun
Not seeing much of a point in the 7700xt if the 7800xt is widely available at MSRP.
It would be interesting at 400 or less
Still nice to have thee reasonable options in this generation
7800XT at 500 for 1440P, 7900XTX at 900 for 4k raster/light ray tracing and 4090 for 4k/ultra/ray tracing
The last gen 6600/6700 still seem better at the lower end than anything current gen
reply
Not seeing much of a point in the 7700xt if the 7800xt is widely available at MSRP.
It would be interesting at 400 or less
Still nice to have thee reasonable options in this generation
7800XT at 500 for 1440P, 7900XTX at 900 for 4k raster/light ray tracing and 4090 for 4k/ultra/ray tracing
The last gen 6600/6700 still seem better at the lower end than anything current gen
reply
Andrew
I am glad GN started doing the overall plus or minus performance relative to another GPU. One constructive criticism I have is uncouple different resolutions on each slide and also put them in order from advantage to disadvantage. The chart seems crowded and hard to read as a snapshot in a video. It provides an information overload that is hard to read at a glance. Love the content!
reply
I am glad GN started doing the overall plus or minus performance relative to another GPU. One constructive criticism I have is uncouple different resolutions on each slide and also put them in order from advantage to disadvantage. The chart seems crowded and hard to read as a snapshot in a video. It provides an information overload that is hard to read at a glance. Love the content!
reply
Add a review, comment
Other channel videos
