• 2 days ago
Sponsor: Montech HyperFlow 360 Cooler on Amazon https://geni.us/dWBIbF6
The AMD Radeon RX 9070 video card is shipping alongside the RX 9070 XT and surprised us with its efficiency. The power-to-performance ratio is strong on the RX 9070 non-XT, seemingly finding a "sweet spot" of balance between power consumption and the output. The 9070 and 9070 XT are known to have a weirdly small $50 gap between them, at an MSRP of $550 and $600, but in either case, the two have ganged-up on the NVIDIA RTX 5070 GPU to effectively invalidate its existence -- even more than NVIDIA already did. This review benchmarks the performance of the RX 9070 vs. the 9070 XT, RTX 5070, RX 6700 XT, 6600, and plenty more.

The best way to support our work is through our store: https://store.gamersnexus.net/

Like our content? Please consider becoming our Patron to support us: http://www.patreon.com/gamersnexus

Watch our 9070 XT review: https://www.youtube.com/watch?v=yP0axVHdP-U
Learn about the AMD RDNA 4 9070 architecture: https://www.youtube.com/watch?v=UAe50byQGG0
Watch our NVIDIA RTX 5070 review here: https://www.youtube.com/watch?v=ntSylZ1Bp1Y
And our review of NVIDIA's 5070 Ti: https://www.youtube.com/watch?v=PhtVic3Vm0Y

RELATED PRODUCTS [Affiliate Links]

AMD RX 9070 on Newegg (when it lists): https://howl.link/zbckuef2rmkuz
AMD RX 9070 on Amazon (when it lists): https://geni.us/m8We7o

AMD RX 9070 XT on Newegg (when it lists): https://howl.link/8kub4h6isbhld
AMD RX 9070 XT on Amazon (when it lists): https://geni.us/Do7nDI

AMD Ryzen 9800X3D CPU on Newegg: https://howl.link/o17puhoclww0v
AMD Ryzen 9800X3D CPU on Amazon: https://geni.us/bsS1kFv

NVIDIA RTX 5070 on Newegg: https://howl.link/n6tq3hig9w8hk

NVIDIA RTX 5090 on Newegg: https://howl.link/mcb343fapkti6
NVIDIA RTX 5090 on Amazon: https://geni.us/CGgM6y

TIMESTAMPS

00:00 - AMD RX 9070 GPU Review
03:16 - Basics, Specs, Differences from 9070 XT
05:46 - Dragon's Dogma 2 4K Benchmarks
06:29 - Dragon's Dogma 2 1440p GPU Comparison
07:30 - Dragon's Dogma 2 1080p Best GPUs 2025
08:08 - FFXIV Dawntrail 4K Benchmarks
09:13 - FFXIV Dawntrail 1440p 9070 vs 5070
09:37 - Resident Evil 4 4K RX 9070 XT vs 9070
10:35 - Resident Evil 4 1440p Best Video Cards
10:59 - Resident Evil 4 1080p Graphics Card Test
11:14 - Black Myth Wukong 4K GPU Differences
11:50 - Black Myth Wukong 1440p Benchmarks
12:31 - Black Myth Wukong 1080p
12:59 - Starfield 4K
13:34 - Starfield 1440p
14:12 - Starfield 1080p
14:25 - Dying Light 2 4K
15:02 - Dying Light 2 1440p
15:26 - Cyberpunk 2077 Phantom Liberty 4K
16:01 - Cyberpunk 2077 Phantom Liberty 1440p
16:20 - Cyberpunk 2077 Phantom Liberty 1080p
16:47 - Ray Tracing: Black Myth Wukong 4K
18:09 - Ray Tracing: Black Myth Wukong 1440p
18:35 - Ray Tracing: Black Myth Wukong 1080p
18:59 - Ray Tracing: Dragon's Dogma 2 4K
19:18 - Ray Tracing: Dragon's Dogma 2 1440p
19:43 - Ray Tracing: Dragon's Dogma 2 1080p
19:59 - Ray Tracing: Dying Light 2 4K
20:15 - Ray Tracing: Dying Light 2 1440p, 1080p
20:44 - Ray T
Transcript
00:00This is the 9070 non-XT. It's supposed to be $550 for MSRP at least, the XT is $600.
00:06So that makes the XT about 9% more expensive.
00:08They are really close to each other for pricing.
00:11It looks kind of like an upsell, but the 9070 actually remains really interesting on its own.
00:17And it's not actually for a reason that we expected.
00:21It's not even because it stands up really well when you put it down that way,
00:24although that is quite impressive.
00:26First of all, in some of our efficiency tests, when looking at performance per watt,
00:28it nearly tied for the most efficient GPU.
00:31So this seems to be in the so-called sweet spot of power and performance
00:34in what would be a huge upset favoring AMD.
00:37The fact that it's starting to chip away at one of NVIDIA's key areas of competition is huge.
00:41The reason for that is simply because the 9070 pulls a lot less power than the 9070XT by percentage,
00:47but most of the performance remains.
00:49And so for the quickest possible version of the performance between these two cards,
00:53the 9070XT versus the 9070 in our testing has the XT model in the range of 9% to 16% better
01:00at 4K and 1440p rasterized, again, against the 9070.
01:04Typically, it's about 11% to 13% in the game suite we did.
01:07I think we did 10 games rasterized, and then we did another suite with RT.
01:11At 1080p, it is 6% to 15% better for the XT versus the non-XT,
01:17mostly around 9% to 11% for the games we tested.
01:20Now, as for the 5070, it's not looking good for NVIDIA.
01:23Not that it looked good for NVIDIA when they got on stage and said this.
01:274090 performance at 549.
01:32That really was the nail in the coffin before the 5070 even launched.
01:37Except this time, Jensen Huan was holding the hammer, I guess.
01:41Anyway, versus NVIDIA's RTX card you shouldn't buy, at 4K and rasterized,
01:45the 9070 leads the 5070 by 0% to 18%, depending on the game.
01:50It has one loss in Final Fantasy, and it has losses in some ray tracing tests,
01:54but not all of them.
01:55That's also interesting.
01:56The 5070 is basically out of the conversation, so at this point,
01:59we're left with a discussion of the 9070 XT versus the 9070,
02:02which is probably what AMD wants.
02:04That is what NVIDIA wants, because NVIDIA has done NVIDIA minus $50
02:09for, like, all of its pricing, where it has a card every $50 to $100.
02:13AMD, meanwhile, has switched from NVIDIA minus $50 to AMD minus $50.
02:21That makes this a different discussion than it might otherwise have been.
02:24This might be partly due to yields.
02:26If they're yielding enough that the XT is readily available
02:29and they want to create an upsell, that might be a motive for it.
02:31We don't really care what their reasoning is.
02:33It's irrelevant to us.
02:35What we care about is, as a consumer, which one should you buy?
02:38And really, it doesn't make any sense to either spend the extra $50
02:42or save the extra $50, depending on which one you're looking at.
02:45So that's all we really care about here.
02:46And that's what we're going to get into today.
02:48So let's get to the review.
02:49Before that, this video is brought to you by the Montech HyperFlow series of liquid coolers.
02:53The Montech HyperFlow aims to compete with other affordable liquid coolers
02:57by fighting on price to performance.
02:59In our testing, cooler performance was comparable to others in its class.
03:02The ARGB HyperFlow has an illuminated pump block with RGB LED fans
03:07that have quality rubber bumpers for vibration damping.
03:10The cooler has a six-year warranty and supports all modern Intel and AMD sockets.
03:14And you can learn more at the link in the description below.
03:16All right, so this one's going to be pretty simple.
03:18We're not going to spend a lot of time on the architecture today
03:20or really the comparisons.
03:21They won't be as long as in the XT review.
03:23So if you want the full sort of discussion of the comparisons,
03:26then go to the 9070 XT review.
03:28We'll link that below.
03:29And if you want the full discussion of the architecture,
03:31so this is things like changes to the actual what's going on in the block diagram,
03:35changes to the ray tracing, stuff like that,
03:37check out our news piece on the RDNA 4 announcement.
03:40That'll be linked below as well.
03:42This one, we're kind of going for the speed run here for a few reasons.
03:44One is we just talked about the 9070 XT.
03:47A lot of that information matches with the 9070 except for the performance.
03:51And the second reason is I haven't gotten to sleep
03:54because there's non-stop hardware chaos or launches.
03:58I mean, it's a good problem to have.
04:00I love working on this stuff.
04:01It's been a really fun week.
04:02But also, I am going to die.
04:05So I need to get through this review and then sleep.
04:07But it's a really interesting card.
04:09The basics architecturally match the XT.
04:12So the 9070 still benefits from the ray tracing changes that AMD has made.
04:16That was the sort of biggest area they focused on in their gaming discussion.
04:19The 9070 has 56 compute units with the 9070 XT running 64.
04:24Both are 16 gigabyte cards with GDDR6 on a 256-bit bus.
04:28The GPUs have slightly different clocks.
04:30The architectural overhauls of RDNA 4 match.
04:33So the same changes apply to the 9070 as the XT again,
04:36with sort of the main exception being that 5060 user versus 64.
04:40That's really sort of it.
04:41It's a very simple setup this time.
04:43Now, pricing, we're not going to bother going through alternative cards right now
04:46because there's really just three that are in the discussion today
04:49for a card in this class of price.
04:51It's going to be the 5070, which you shouldn't buy,
04:54the 9070 XT, and the 9070.
04:57And then Intel, they're starting to get closer, but not for the right reason.
05:02So the B580 is supposed to be a $250 card, maybe $300 with a partner model.
05:06It's in like the $400 sometimes, and you shouldn't buy it for that price, unfortunately.
05:11Very interesting card at $250, though.
05:12They're super competitive, actually, at that price.
05:15But anyway, we're going to get straight into the data.
05:17We'll save some of the pricing recap and alternatives
05:20and comparing all the cards available after this launch clears.
05:24So we're going to give it a couple days, come back and revisit post-launch,
05:27just because we want to monitor the next day or two,
05:29get an idea for the availability of both the 5070 and the 9070 series,
05:33and then look at the pricing once it kind of hits the street
05:36so we get the actual street pricing.
05:38But for now, we have to work off of MSRP for comparisons
05:40because the only fixed comparison we have until we can see what this price is at.
05:45Let's get straight into the data.
05:46Dragon Stock with two is up first and at 4K initially.
05:48In this test, the RX 9070 non-XT ran at 64 FPS average with lows at 54 and 53.
05:54The 9070 XT's 70 FPS result positions at 9% ahead,
05:58which is exactly how much higher the base MSRP is.
06:01Against the equivalently priced RTX 5070, this is one of AMD's stronger titles.
06:07The 9070 Pulse leads the 5070 FE's 56 FPS average by 14% here,
06:11with 1% and 0.1% lows improved in step with the average.
06:14The 9070 is achieving 87% of the performance of the 5070 Ti,
06:18but at 73% of the base MSRP.
06:21The RX 9070 leads the RX 7800 XT by 37% here as well,
06:26and we'll look at the 6700 XT elsewhere.
06:29At 1440p, the RX 9070 ran at 106 FPS average,
06:32with the 9070 XT leading it at 116, or 9.3% ahead.
06:37The 5070 Ti is only a few FPS higher than the 9070 XT,
06:41with the 4080 Super just beyond that.
06:43In short, the 9070 is achieving 88% of the 4080 Super's performance,
06:49which was a $1,000 card, but the 9070 is at 55% of the original MSRP.
06:56Despite our wishes that the 9070 were $50 cheaper,
07:00this at least helps to slowly reset the runaway pricing that Nvidia set in motion.
07:06It gives everyone some perspective of,
07:08no, we should not be paying $1,000 for that.
07:11For some comparisons against prior generations,
07:14the 9070 improves on the 67 FPS result of the 7700 XT by 59%,
07:19the 53 FPS result of the 6700 XT, which was a very popular card for them,
07:25by 100%, and the 42 FPS of the 6600 XT by 152%.
07:30Here's 1080p.
07:31The older GTX 1060 and 1070 also appear on this chart to give you some insight into upgrades.
07:36Although anything modern would be better than those,
07:38it's still helpful to see them on the charts.
07:41There's no point in putting percentages in for them though,
07:43because they go from unplayable to playable.
07:46The RX 9070 is about tied with the 7800 XT.
07:48The 9070 XT still leads by around 8%, with the 5070 Ti ahead of the 9070 by 13%.
07:56The 9070 leads the RTX 5070 DOA edition by 6%,
08:00and the lead over the 6700 XT is 96%,
08:03with the improvement on the 6600 non-XT's 49 FPS result at 173%.
08:08Final Fantasy XIV Dawn Trail is up now.
08:10We included this same disclaimer and disclosure in our 9070 XT review,
08:14but to recap it, we've noticed lower performance for the 9070 series specifically
08:19in Final Fantasy XIV Dawn Trail, including lower results than AMD's own claims.
08:24We've had highly reliable results in this benchmark,
08:26and our 9070 series data has been repeatable across multiple tests.
08:30So it doesn't really matter what the first party data from its presentation says,
08:34we're publishing it.
08:35But we just want to highlight that there is a disproportionately low performance for the 9070s,
08:40and that's something that we're investigating.
08:41It's possible that it could be the specific scene,
08:43maybe the benchmark version, or just the architecture's behavior here naturally,
08:46but the result is repeatable, and so we're running them.
08:49At 4K, the 9070 continues the scaling pattern set by the 9070 XT.
08:53The XT is again 9% ahead.
08:55The 5070 FE leads both of the cards in this test,
08:59running at 78 FPS average and improving upon the 9070's result of 63 FPS by 24%.
09:05Against the prior generation,
09:07the 9070 XT sees a smaller improvement on the 6700 XT of 53%.
09:13At 1440p, the 9070 ran at 126 FPS average.
09:16The 9070 XT is 10% ahead here, with the 5070 ahead by 21%.
09:20Improvement over the last generation is non-existent,
09:23with the 7800 XT right next to the 9070.
09:25Now as we said in the XT review,
09:27we checked in with Der Bauer and saw similar results to some of his tests,
09:31so it does appear to be a behavior that is
09:34related to the game engine and the API and things like that.
09:37Moving on, Resident Evil 4 is up next, tested at 4K first.
09:40The 9070 ran at 91 FPS average here,
09:42creating a 13% lead for the 9070 XT and its 103 FPS result.
09:46The 9070 XT was excitingly right next to the 4080 Super's 105 FPS in this game,
09:52meaning that the same $1,000 4080 Super result
09:56leads the $550 MSRP 9070 by only 16%.
10:02That is a paltry gain for such a huge price difference,
10:05especially given how new the 4080 Super is.
10:08AMD's 9070 exceeds the performance of the same MSRP RTX 5070's 78 FPS by 16%.
10:15If Nvidia would stop remaking 4070 Supers and 4080s,
10:19they might actually make some progress here.
10:22But until they do, the memes will continue until the products improve.
10:27And now I realize it's too much.
10:28Now, yes, the only thing that's too much, Jensen,
10:31is the amount of 4070 Supers and 4080s.
10:35At 1440p, the RX 9070 ran at 172 FPS average and outperformed the 3090 Ti.
10:39The 9070 XT outperforms the 9070 by 13%,
10:43with the 5070 Ti only slightly beyond that result.
10:46Against the 5070's 152 FPS average, the 9070 runs 13% ahead.
10:50The 5070 really does look even more DOA than previously in some of these tests,
10:55if that were possible.
10:56If it were a dead horse, it'd be getting kicked right now.
10:591080p is up now.
11:00The 9070 ran at 247 FPS average here,
11:02which had it between the 4070 Ti and the 4070 Ti Super.
11:05The 5070 still trails, now giving the 9070 a lead of 10%.
11:10The XT model maintains about a 12% advantage over the non-XT.
11:14Black Myth Wukong is a much more challenging title for AMD to run,
11:17but mostly with ray tracing.
11:19Either way, there should be some more favor for Nvidia here.
11:22Even without RT, it's a heavy game.
11:24The ray tracing test will come a little bit later,
11:26but let's look at 4K and rasterized.
11:28The 9070 ran at 41 FPS average against the 9070 XT's 46,
11:33giving the latter a lead of 11%.
11:36The 7800 XTX sits just ahead of the 9070 XT.
11:39The 9070 ends up still leading the RTX 5070, albeit on a technicality,
11:44which is actually a huge victory for AMD in this particular title.
11:47That'll change with RT,
11:48but this is a good position overall for raster.
11:50At 1440p, the 9070 held 75 FPS average with well-paced lows,
11:54sitting at 65 and 60 FPS for 1% and 0.1%.
11:58This is consistent and good for frametime pacing,
12:01but it's also about the same as all of its neighbors,
12:03so it's nothing exceptional.
12:04The 5070 ran at 72.1 FPS average.
12:07The difference between the two cards is imperceptible
12:10and shows up in measurements,
12:11but would not be noticeable to an end user.
12:14The 9070 XT's 83 FPS result has it again about 11% ahead of the 9070,
12:18so that's pretty predictable now.
12:20To just highlight some landmarks,
12:22the 2060 ran at 25 FPS and tied the RX 6600,
12:26the 6600 XT ran at 29,
12:28and the 3060 ran at 31 FPS average.
12:31At 1080p, the 9070's 104 FPS average allowed it
12:34to finally start gaining some distance on the 5070.
12:37It's not a meaningful swing at 6% improved,
12:39but is in favor of the 9070.
12:41If AMD were $50 cheaper, it'd be a very clear victory.
12:45Unfortunately for AMD,
12:47a tie will often still be decided in NVIDIA's favor
12:50by most buyers, especially mainstream,
12:52but also because of the complications of RT and software suites.
12:56But still, beating the 5070 in this game is an achievement for the 9070.
12:59Starfield is up now, tested at 4K.
13:01In this one, the RX 9070 ran at 63 FPS average
13:04and roughly tied the RTX 4070 Ti Super.
13:07The 9070 outperforms the RTX 5070's 54 FPS average by 17%,
13:12which is starting to be wide enough
13:13that NVIDIA needs to be concerned,
13:15if they weren't basically a monopoly,
13:18but you never know when the first domino falls.
13:20The 5070 Ti falls below the 9070 XT in this one
13:23and only leads the non-XT by about 6 FPS,
13:27hardly enough to spend $200 more for.
13:30The proximity to the 4080 and 4080 Super
13:32also isn't great news for NVIDIA.
13:34At 1440p, the 9070 ran at 96 FPS average,
13:37and nearly tied the 4070 Ti Super again.
13:39The 5070 Ti outperforms the 9070 by a lame 5.4%,
13:44with the 9070 XT outperforming the 9070 by 10%.
13:48Against the 5070's 83 FPS average,
13:51the 9070 leads it by 15%.
13:53If the 5070 weren't already in the grave,
13:56we would pronounce its impending demise.
13:58We might want the 9070 to be a bit cheaper,
14:01just to see that balance for the times it loses,
14:04or is tied, like in the next game.
14:06But even without that,
14:07we can't deny that AMD is putting up
14:09its best fight against NVIDIA in years.
14:11At 1080p, the 9070 runs at 118 FPS average.
14:14The relative rank hasn't moved much.
14:16It leads the 5070 by 14% here,
14:19so the lead has slightly reduced,
14:21but it's still ahead.
14:22The 9070 XT leads the 9070 by 8%.
14:25In Dying Light 2 at 4K,
14:26the 9070 ran at 56 FPS average
14:29and kept lows remarkably close by,
14:31at 50 and 41,
14:32but not in any more remarkable way
14:34than all of its neighbors.
14:35The game is just consistent.
14:36The 9070 ends up tied with the 5070.
14:38There's no difference between those two
14:40and in these situations,
14:42NVIDIA will generally be viewed
14:43more favorably by a mainstream audience.
14:45That's where the price would come in,
14:47or hopefully the awareness of other results.
14:49The 9070 XT leads the 9070 by 11%,
14:51so you'd be paying 9% more at MSRP
14:53for about 11% more performance.
14:55This is one of those games
14:56where the 5070 Ti pulled ahead
14:57of the 9070 XT in a noteworthy way,
15:00so we see some of that apply
15:01to the 9070 and the 5070.
15:02At 1440p, the 9070 runs at 106 FPS average
15:06and ties the 5070 again.
15:07They're within error of each other.
15:09The 9070 XT outperforms the 9070 by 11%,
15:12with the 5070 Ti still ahead of the XT.
15:14Generationally,
15:15we'll just briefly highlight
15:16the RX 6600,
15:18the 2060,
15:19and the RTX 3060
15:20as reference points
15:21to some of the most common GPUs
15:23of past generations.
15:24Those should give you an idea
15:25to the performance gains.
15:26Cyberpunk is up now.
15:27This one has generally been tough
15:29for AMD in the past,
15:30but it's harder for it with ray tracing.
15:31We'll start with rasterized testing at 4K.
15:33The results are thus far positive.
15:35We already knew that the 9070 XT
15:37outmatched the 5070 Ti,
15:39but now the 9070 is nearly tying the 4080.
15:42That's just embarrassing
15:43for Nvidia's last-gen,
15:46I don't know what you call it,
15:47mid-range-posed-as-a-false-flagship GPU
15:50as compared to a $550 video card.
15:53The 9070 technically outperforms
15:55the 7900 XT as well,
15:56though is basically tied,
15:58and the lead over the 5070 is 15% again.
16:00At 1440p, the 9070 now falls
16:03slightly below the 7900 XT,
16:04so they've traded places
16:05as resolution came down,
16:07and the 5070 has also gotten closer.
16:09The 9070 is now about 11% ahead
16:11of the 5070 rather than 15% before.
16:14The 9070 XT leads the 9070 by 11%,
16:17and the 5070 Ti slightly.
16:19In Phantom Liberty at 1080p now,
16:21the 9070 landed at 148 FPS average,
16:23allowing the 7900 XT
16:25to maintain its slight lead.
16:26The improvement in the 9070
16:28over the 5070 is now down to 7%,
16:30from 15% at 4K originally.
16:33If we dropped it down to 360p,
16:35based on this trend,
16:36Nvidia might actually pull ahead.
16:38Like, unironically, I did that math.
16:41That's about where they should cross over.
16:43So great, it's just like they've wanted.
16:45Maybe that's the push for DLSS.
16:47Ray Tracing is up next.
16:48This is where AMD has invested
16:50heavy effort with its architecture,
16:51but all of it has been to try and catch up.
16:54Nvidia was so far ahead in some games
16:56that AMD just won't be able
16:58to catch them in all scenarios.
16:59The real question is,
17:00how much closer do they get this time?
17:02We'll start with AMD's worst case,
17:03which is Black Myth Wukong,
17:05but now with Ray Tracing
17:06and at 4K with upscaling.
17:08The 9070 ran at 26 FPS average here,
17:10so outdoing the 3070 Ti,
17:12but below the 3080 from 2020.
17:14The 3080's original MSRP was around $700,
17:18although that really only held
17:19for a week or two
17:20before COVID-era scalping went crazy.
17:22The 9070 XT outdoes the 9070 by 14% here.
17:26This is too low of a frame rate
17:28to be playable,
17:28but we do the test for relative scaling
17:30and percent scaling.
17:31The 5070 crushes the 9070 here,
17:34holding a 57% lead over the AMD GPU.
17:36The 5070 Ti's scaling over the 9070 XT
17:39was better at nearly 80%,
17:41but 57% is still huge
17:44for the 5070 over the 9070.
17:45If you want to play Black Myth
17:47with Ray Tracing,
17:48like we said last review,
17:49you basically only buy Nvidia.
17:51Lows are where Nvidia will struggle the most
17:53for some scenarios
17:54with the 5070 VRAM capacity.
17:56In this particular test,
17:57we're not hitting VRAM saturation
17:58during the testing.
17:59However, we have shown instances
18:01where the 5070 hits saturation
18:02with heavy workloads
18:03and high resolutions
18:04leading to heavy stutter,
18:06like in the Cyberpunk RT 4K scenario yesterday.
18:09At 1440p upscaled,
18:10the 9070 pushed 47 FPS.
18:12This gave the 9070 XT about a 6 FPS lead,
18:15nearly tied the 3080,
18:17and outperformed the 3070 Ti
18:19and 7900 XTX notably.
18:20The improvement over the 7900 XTX
18:22shows that AMD has actually executed
18:25on its RT performance uplifts.
18:27But unfortunately for AMD,
18:29the 5070 remains superior
18:30with the 73 FPS average.
18:32It leads the 9070 by 55%.
18:35That's huge.
18:35At 1080p,
18:36the 9070 measured at 67 FPS average
18:39and pushed past the 3080.
18:41The 9070 XT leads by 9%.
18:43The 5070 leads the 9070 by 46%.
18:47So it has slightly reduced,
18:49but still an unbeatable lead.
18:51Dragon's Dogma 2 with Ray Tracing is next,
18:53tested at 4K first.
18:54We would play Dragon's Dogma
18:56with Ray Tracing for this lineup.
18:58It's not as heavy.
18:59We measured the 9070 at 56 FPS average,
19:02giving the 9070 XT a lead of 9%.
19:04The 4080 isn't much past that,
19:06so for as close as these cards are,
19:08they are also very close
19:10to the 4080 and the 4080 Super.
19:13The 5070 trails in this one,
19:15giving the 9070 a lead of 15%.
19:17At 1440p,
19:18the 9070 gets dangerously close
19:20to Nvidia's 5070 Ti,
19:22which failed to pass the 9070 XT
19:24in the previous review.
19:25The 5070 Ti leads the 9070 by 9%,
19:28the same lead as the 9070 XT has against it.
19:31The 5070 non-Ti is down at 83 FPS average,
19:34creating a 13% lead for the 9070.
19:37This chart has entries for the 6700 XT
19:40and other prior GPUs,
19:41so if you want to pause for those,
19:42you can check them out.
19:43At 1080p,
19:44the 9070 ran at 119 FPS average
19:46and led the 7900 XT by a few frames per second.
19:49The lead over the 5070's 107 FPS average
19:51was 11%.
19:52The 9070 XT has a reduced lead
19:54over the non-XT at this resolution,
19:56now just 6% ahead of the cheaper card.
19:59Dying Light 2 with upscaling at 4K
20:01is up now,
20:01this time with ray tracing.
20:02The 9070 XT leads the 9070
20:04by a higher than average lead of 16% here.
20:07The 9070 ends up below the 5070 in this one also,
20:11with the 5070 producing a lead of 9%
20:13over AMD's most relevant card.
20:15At 1440p,
20:16the 9070 runs at 78 FPS average
20:18and gives the 9070 XT a lead
20:20of about 10 to 11 FPS.
20:22The 9070 XT's prior 16% lead
20:25is now reduced to 14%.
20:27The 5070 is also reduced to a 4% lead
20:30from a 9% lead at 4K.
20:32The 9070 was really struggling
20:34with that 4K resolution though.
20:35Finally for this game,
20:361080p puts the 5070 at just 1 FPS ahead now,
20:39so basically margin of error.
20:41The 9070 XT leads the 9070 by 13%.
20:44In Resident Evil 4 with ray tracing
20:45and at 4K upscaled,
20:47the 9070 held a 104 FPS average
20:49and outperformed the 4070 Ti Super and 5070.
20:52The boost over the 5070 is 14% here.
20:55The 9070 XT ran 12% higher framerate
20:58than the non-XT,
20:59with the 5070 Ti basically tied with that.
21:01This chart has a lot of cards
21:03lower down the stack,
21:04like the 2060, 3060, 2070,
21:06and 6700 XT for past generation comparisons.
21:09At 1440p upscaled,
21:10the 9070 ran at 162 FPS average
21:13and roughly tied the prior 7900 XT.
21:15The 9070 XT was 11% ahead of the non-XT here,
21:19with the non-XT ahead of the 5070 by 9%.
21:22Cyberpunk is up last,
21:23one of the two heavier titles we use for RT.
21:26We test this one at two groups of settings,
21:28Ultra and Medium,
21:29chosen because Nvidia gains
21:31a disproportionate advantage at Ultra.
21:33Medium, meanwhile, pushes them closer together.
21:35That gives us the full picture.
21:37At 4K and RT Ultra,
21:38just to stress test it for relative scaling,
21:40the 9070 ran at 18 FPS.
21:41That's obviously unplayable,
21:43but this makes the point we made in our 5070 review,
21:45which is that there are scenarios
21:47where the 0.1% lows on the 5070 can suffer in a big way.
21:51We pointed this out in the last two reviews,
21:53but that's because they're indicating to us
21:55that underneath those lows,
21:56which really are just meant to be an indicator,
21:58they're not the problem,
22:00they point to the problem,
22:01which is spiky and erratic frame time performance.
22:04This is caused by exceeding VRAM limitations on the 5070,
22:07which simply isn't equipped enough
22:09to deal with these problems.
22:10At 4K Medium, the 9070 boosted up to 24 FPS,
22:13no change in how playable it is or isn't,
22:15but even this lighter weight workload is too much
22:17for the 5070 FE's capacity,
22:19which you can see in its lows again.
22:20The 9070 XT leads the 9070 by 15% here,
22:24one of the higher percent increases.
22:25Reducing the resolution helps reach playable frame rates.
22:27The 1080p RT Ultra result
22:30has the 9070 at 61 FPS average
22:32against the 9070 XT's 71.
22:34That's a relatively large advantage
22:36as a percentage of 16% for the 9070 XT.
22:39The 5070 ran at 64 FPS here,
22:42outdoing the 9070 by just 5.7%.
22:44Finally, the Cyberpunk 1080p Medium results
22:47put the 9070 at 81 FPS average,
22:49roughly tying the 7900 XT X.
22:51This also puts it above the 3080.
22:53The 5070 ran at 81.9 FPS average
22:56and lows were within error of the 9070.
22:59There's no distinct advantage in lows for the 9070 here,
23:02average is the same,
23:03which is a major positive for AMD
23:04when considering its previous RT deficit.
23:07We're moving on to power and efficiency testing now,
23:09which looks at both the total power consumption
23:10of the device during the test and the frame rate.
23:13The end result is that we can calculate
23:14the efficiency and performance per watt.
23:16It's actually a really exciting test
23:17for the 9070 specifically.
23:19An F124 at 4K and with RT,
23:22AMD has exceptional performance.
23:24These are only showing two decimals,
23:26but more are processed for the bar width.
23:28The 9070 roughly tied the 4080 Super and 5090 FE
23:32as the most efficient GPU in the suite here.
23:34The 5070 was at 0.17 FPS per watt,
23:37giving the 9070 an advantage of 24%.
23:40It's a huge swing in its favor.
23:42The 9070 XT was closer to the 5070 here.
23:45At 1080p but with the same settings and game,
23:46the 9070 again ends up just behind first place.
23:49It's tied with the 5070 Ti
23:51and just behind the 0.64 result of the 5080.
23:54The 9070 XT burns more power
23:56and although its performance is better,
23:58the trade-off favors the 9070 here.
24:00The 5070 ends up way down the ranks at 0.55 here.
24:03Final Fantasy 4K is up next.
24:05This is one of AMD's weaker showings
24:06given its limited performance scaling in Final Fantasy XIV.
24:09At 4K, the 9070 pulled 224 watts during this test
24:12when measured at the PCIe cables and the slot.
24:15The 5070 pulled 10 watts more
24:17but had better efficiency from its higher performance.
24:19The end result is that the 9070 outdid the 7900 XT for efficiency
24:24but gave the 5070 a lead of about 18%.
24:27For reference, the 9070 XT pulled 310 watts here,
24:31roughly matching their TDPs.
24:32At 1440p, the 9070 ranked at 0.56 FPS per watt
24:36with a total 224 watt draw.
24:38That has the 4060 as more efficient,
24:40the B570 as slightly less efficient,
24:42and the 7900 XT as slightly less efficient also.
24:45The NVIDIA 5070's 0.65 FPS per watt result
24:48outdoes the 9070 again here,
24:50but we'll see if that persists
24:52outside of AMD's weakest rasterization title.
24:54Dragon Stock with 2 with RT
24:56shows exceptional efficiency for the 9070.
24:58At 224 watts again and 0.42 FPS per watt,
25:02it's looking pretty good overall.
25:03The performance is slightly lower than that of the 9070 XT,
25:06but the reduction in power really puts it in a better spot
25:09on the efficiency curve.
25:10It seems like this is more of the so-called
25:13sweet spot for power to performance.
25:15The 5070 is down at 0.36 FPS per watt
25:18despite pulling 231 watts.
25:19In Starfield at 1440p,
25:21the 9070 is less impressive than an F1
25:24and Dragon Stock with 2,
25:25but still overall good.
25:27Its result was 0.43 FPS per watt,
25:29a large improvement on the 9070 XT's 0.34.
25:33The 9070 ties the 5070,
25:35which is a problem for NVIDIA,
25:36because efficiency was one of NVIDIA's main benefits
25:39with the 5070 Ti against the 9070 XT.
25:41So if AMD is starting to chip that away,
25:44that's bad news for NVIDIA.
25:45That's one of the key things that they've kind of stood upon
25:48for being more expensive in the past.
25:49That's it then.
25:50First of all, credit to the 9070
25:51for being one of the most power efficient cards
25:53on our charts now.
25:54It's like basically tied at the top
25:55in a few of those charts, not all of them,
25:57but it's looking better there.
25:59Now in our 9070 XT review,
26:01we said that AMD made the right choice
26:03for pushing for higher performance on the XT
26:07rather than pulling back some of its power budget
26:10to get better efficiency.
26:11So that was kind of the trade-off
26:12that any silicon manufacturer has.
26:14They can choose efficiency with lower performance
26:18or they really push the power envelope
26:19and get kind of past the peak point on the efficiency curve.
26:23So the XT goes past that point
26:25and the 9070 non-XT pulls it back.
26:28It looks really good on the charts,
26:29which is actually really important for AMD
26:32because that is one of the things on the bullet list
26:34of reasons you buy NVIDIA without checking performance
26:37if you're sort of a mainstream buyer.
26:39And they're chipping away at that.
26:41Now the reasoning, the difference is so big against the XT
26:45is because it seems that the 9070 XT
26:48just pushes past the peak of that curve
26:51to maintain the clocks and that CU count necessary
26:54in order to be actually pretty competitive.
26:56Whereas the 9070, it gets most of the benefit
26:59of the architecture
27:00and it has a slightly reduced CU count.
27:02They can pull the power drawback
27:04and AMD also gets the benefit from
27:06because they've reduced the power draw
27:07by 80 watts on the 9070 for out of box,
27:11they can create more of a gap
27:12between their own XT and their non-XT.
27:14So they're kind of playing it from both angles,
27:16which is what they need to do.
27:18The 9070 XT ends up ahead of the 9070 by about nine to 12%
27:22in most of the games we tested
27:24with the 5070 generally giving the 9070 a lead
27:27in several games.
27:28Dragon's Dogma 2 at 4K was playable on both,
27:31but the 9070 had about a 14% advantage over the 5070.
27:34The 9070 also led the 5070 by about 15%
27:37in Cyberpunk rasterized at 4K
27:40with the 9070 XT ahead of the 9070 by 12%.
27:43There remains one breakout for the 5070 in Final Fantasy,
27:46something we talked about in our XT review,
27:48but beyond that for rasterization,
27:50the 9070 is ahead in basically every scenario.
27:52Baldur's Gate 3 even showed a large benefit at 4K
27:55for the 9070 versus the 5070
27:57since we're far enough away from the CPU bottleneck.
28:00In ray tracing at 4K,
28:01the 9070 XT ended up around 9% to 19%
28:04a higher frame rate than the 9070,
28:06though that 19% number is based on a low frame rate,
28:08but even still 17 is the next highest percentage.
28:111440p and 1080p weren't that different
28:13and can be found in the chart section earlier.
28:15As for the 5070 in RT and 4K,
28:18the 9070 beat it by 15% in Dragon's Dogma 2,
28:218% in F1-24, 6 to 7% in Cyberpunk,
28:2514% in Resident Evil 4.
28:26Now these summary charts, as a reminder,
28:28these are calculated as percent improvement
28:29from the 5070 to the 9070.
28:32The thing to be aware of is that
28:33because the math is calculated as a percent improvement
28:36from 5070 to 9070,
28:38those negative numbers are sort of going
28:40in the opposite direction.
28:41So the really proper way to do this
28:42would be to calculate them all individually
28:44as percent improvement for the highest minus the lowest,
28:47but we just spent 30 minutes doing that
28:48so you can find that in the rest of the charts earlier.
28:51Back to the percentages,
28:52this is the reason we don't like
28:53these summary charts generally,
28:54but the Black Myth result,
28:56if you were to calculate it in the same direction,
28:57then it ends up being 57% improved
29:00for the 5070 over the 9070.
29:03So that would be the more comparable number
29:05to the rest of the positive numbers we're looking at.
29:07Now the 9070 lost in Black Myth and Dying Light 2 with RT.
29:11At 1440p and RT, the same losses exist.
29:14So this puts the non-XT and the XT
29:16in a similar position to what we saw
29:17with the 7800XT and the 7700XT
29:20or the 7900XT and the 7900XTX.
29:23The gap is small enough that most people
29:25will basically upsell themselves
29:27to the XT model in this case.
29:29Weirdly, NVIDIA's own fumbles
29:31have sort of made the proximity of these two cards
29:33less of a problem than it might otherwise have been
29:36because from AMD's perspective,
29:38it's AMD or AMD.
29:40They probably don't really care that much
29:41which of AMD you buy,
29:43but the XT makes more sense than a 5070 does,
29:46assuming the MSRP for both.
29:47The 5070 is already mostly irrelevant.
29:50NVIDIA's strengths of its DLSS,
29:52Ray Reconstruction Reflex,
29:53all those remain, of course,
29:55but the bells and whistles just,
29:56we don't think are good enough on the 5070
29:59to outweigh the 9070 or the XT.
30:01The biggest strength NVIDIA has is CUDA.
30:04So for applications where you need CUDA acceleration,
30:06the software we use for a lot of the production work we do,
30:09it's all CUDA accelerated.
30:10So it doesn't really matter.
30:12We're in for the 4090s or the 3090s or whatever
30:16because that's just what makes our work go faster
30:18for things like video editing.
30:20But from a gaming standpoint,
30:22the 9070 and the 9070 XT have really kicked the ass
30:26of the 5070, to be fair though.
30:29The 5070, it was like,
30:31it was kind of off to the side,
30:32kicking its own ass.
30:34And then the 9070s came in and they're like,
30:36oh, that looks fun.
30:37And they joined in.
30:38I want to thank all of you for coming today.
30:40So, I don't know.
30:42Anyway, that's it for this one.
30:44Thanks for watching.
30:45Subscribe.
30:45We're going to talk more about the pricing
30:47over the next few days
30:48as we get a chance to see it unfold.
30:49So might be hardware news,
30:51might be standalone.
30:51I don't know yet.
30:52But yeah, that's going to be it.
30:54I'm going to go to sleep.
30:55Thanks for watching.
30:56To directly support our reviews
30:58and the huge amount of effort it is
30:59to push through this many this quickly,
31:01plus all the other chaos,
31:03go to store.gamertexas.net.
31:04You can help us out directly with our reporting.
31:07All the work we do these days
31:08is majority funded by the audience
31:10and from your support on both the store and Patreon.
31:13And on the store,
31:13you can get some cool stuff
31:14like PC building anti-static mod mats.
31:16You can also grab a project and soldering mat
31:19or one of our copper plated stainless steel mule mugs.
31:22Our coach pack is pretty cool too.
31:23Check it out on the store.
31:24There's a lot of cool stuff there.
31:26Dice kits too.
31:26Thanks for watching.
31:27Thanks for the support.
31:28It's been fun to work on something this cycle
31:29that's not just a complete train wreck.
31:33So we'll have some more stuff coming up soon
31:37and we'll see you all next time.

Recommended