Sapphire Radeon RX 9070 XT Nitro+

[AusWolf]

So it is a cheaper 7900 XTX with better RT and FSR 4. Sweet!

 

[alwayssts]

We all know Silent Hill is a problematic title, that makes it an outlier in itself.

A problematic outlier you include to note averages and minimum placement. That is accurate. That is what we are saying.

One thing I have to add that ER is actually a hard 60 fps capped game and requires a third party software to unlock the framerate (you can't even use that with online services and anticheat) . And pretty much anything over 3070TI can do 60fps even at 4K. So those results doesn't even reflect actual use case for most players either way. Considering their engine has also no hope to be a use in any other title too (Nightreign is just ER, looks and performs same as we saw in network test), I don't really think it should be used for GPU reviews anymore as it can skew the results without representing actual performance you'd get normally. It wouldnt matter if it performed about average but it's just a big outlier in benchmarks.

I didn't know that, although I should have (I played Dark Souls that obviously had similar issues across the series). Thanks for sharing the info, and once again, kind of proving my point.

So it is a cheaper 7900 XTX with better RT and FSR 4. Sweet!

It isn't really even much better RT...kinda more-so similarish. Which is weird, but true. 9070 xt is just clocked higher. You can overclock a xtx to have decent RT...but you may never have FSR4.

It's so messy I just think of it as "If you know what a xtx gives you, and it helps your use-cases that's great, there are reasons for it, but it makes more sense to recommend this now".

Basically, do you want that compute/ram, or FSR4. Pick your poison. Also, this is cheaper.

Again, AMD plays these games too. It's just, you know, what nvidia does is egregious. What AMD might do wrt to FSR4 is just kinda shitty, bc people do eventually need cards with these features across the stack.

It's one way to get people to migrate.

If they ported FSR4 to N31 it *could* potentially hurt this product as XTX has potentially something like 25% (33% at 2631mhz stock vs 2970mhz stock) more compute (depending on if OCing) and 50% more ram.

 

[SilentPeace]

I want AMD to have a win here as much as anyone, but people are calling the win too early. We have learned from prior launches that we need availability and actual MSRP cards to be restocked. Hardware Unboxed is suggesting the $599.99 is a temporary early deal that's going to disappear and eventually the XT will be around $800 or more. If that happens, nothing is won.

Nothing will even be different.

AMD really should have made reference cards.

OcUK confirmed that when their initial supply has sold that the prices will be going up.

 

[AusWolf]

It isn't really even much better RT...kinda more-so similarish. Which is weird, but true. 9070 xt is just clocked higher. You can overclock a xtx to have decent RT...but you may never have FSR4.

How is it not better in RT?

 

[Dr. Dro]

What the hell is going on with Alan Wake 2? That didn't strike me as an AMD title.

Considered Alan Wake 2 performed poorly in earlier iterations of RDNA (especially RDNA 1), it's more than plausible that AMD looked at their floorplans and decided to target the areas where these games demanded and their architecture faltered. So they fixed it, to great result. This is what I like about RDNA 4, it seems they've listened to feedback and worked to improve the areas that their previous generation cards fell short.

A problematic outlier you include to note averages and minimum placement. That is accurate. That is what we are saying.

If you remove it, the 4080 would still come out on top when averaged out. The fact that it's not quite beating the 4080 in an overall scenario does not make it an inferior quality product, nor does it discredit AMD's gains. What it does, however, is exemplify why Nvidia decided to rest on its laurels. And much like what happened to Intel, I hope it comes to bite them in the arse.

The "performance segment", as these cards on the upper-midrange are called, has been utterly stagnated since the release of Ada, and to 7900 XT or RTX 3090 Ti owners, the only things this card truly offers are the improved encoder and better power consumption, and not by much. Hardly worth ~$700. RX 9070 XT is a great product in my eyes. It's the most solid midranger card we've had in a long time, arguably since the RX 480 (and its rebrands). But it is not a revolutionary product.

 

[BlaezaLite]

As good as the 9070XT is, my red devil gre is enough for me. Not enough percentage gain.

 

[AusWolf]

OcUK confirmed that when their initial supply has sold that the prices will be going up.

I hope that's not true. At the current listed price, it's an instant buy for me.

Edit: Checking the OCUK web page, they only talk about pre-release placeholder prices. Nowhere does it say that prices will rise.

 

[alwayssts]

How is it not better in RT?
View attachment 388047

Because 7900xtx has a 2397mhz RT clockspeed. 9070 xt runs RT at 3ghz. You can overclock a 7900xtx to ~3100-3200mhz, you can overclock a 9070 xt to "it's complicated" (and I'm figuring that out on the side.)

Glad you asked, I bet a lot of people didn't know that.

 

[SilentPeace]

I hope that's not true. At the current listed price, it's an instant buy for me.

Edit: Checking the OCUK web page, they only talk about pre-release placeholder prices. Nowhere does it say that prices will rise.

From their forum:

 

[Ruru]

What's with the aweful 30C hotspot difference on the ASUS model? Trash coldplate?

Reminds me of my 6700 XT even with an aftermarket cooler.

 

[AusWolf]

Because 7900xtx has a 2397mhz RT clockspeed. 9070 xt runs RT at 3ghz. You can overclock a 7900xtx to ~3100-3200mhz, you can overclock a 9070 xt to "it's complicated" (and I'm figuring that out on the side.)

Glad you asked, I bet a lot of people didn't know that.

Did you check the percentage I circled? The 9070 XT loses less performance when you turn on RT at the same clock speed. It's something that we haven't seen on Nvidia since Turing, despite their "big improvements".

 

[btk2k2]

I hope that's not true. At the current listed price, it's an instant buy for me.

Edit: Checking the OCUK web page, they only talk about pre-release placeholder prices. Nowhere does it say that prices will rise.

Gibbo says they have hundreds of cards at msrp but when they sell those the price will increase.

Check the ocuk forums.

 

[AusWolf]

From their forum:

View attachment 388048

Dang! Thanks for the info.

 

[alwayssts]

Did you check the percentage I circled? The 9070 XT loses less performance when you turn on RT at the same clock speed. It's something that we haven't seen on Nvidia since Turing, despite their "big improvements".

I'll look into it after I get some rest.

I'm sure there are improvements, and again all of this is solvable (because it's just units and math), but I don't know the equation right now.

I need to really dive into the BVH improvements and all that, still. First day, still.

 

[dartuil]

Gibbo says they have hundreds of cards at msrp but when they sell those the price will increase.

Check the ocuk forums.

its a shame to increase price after the first wave of buyers

 

[LabRat 891]

So it is a cheaper 7900 XTX with better RT and FSR 4. Sweet!

It's *very* close to a 7900 XTX in raster. Notably better in RT.

Even as a 'sidegrade' (ala my RX580X 8GB->RX6500XT 4GB) from my 7900XTX, I'll seriously consider a 9070 XT (if only to get my Gen4x4 slot back )

its a shame to increase price after the first wave of buyers

It sucks, yes. However, that's just simple economics.

I bought my 7900 XTX 'at a premium', and w/in the week it was up another $300-500+ (If you could find one)

 

[Rover4444]

Are the relative performance percentages in the GPU database accurate? It's not looking good to me, performance is similar to a 7900 XT and the same is true in Linux, too. Prices are probably going to skyrocket after current stock sells out as well. Am I the crazy one?

 

[N/A]

?

DB is more of a scripted estimate. tariffs will affect the next batch.

 

[tfp]

1% is not stuttering.

Frame timings would really help in regards to seeing if there is stuttering but it's a different benchmark. TR used to do that back before it was all crypto trash.

 

[LabRat 891]

Are the relative performance percentages in the GPU database accurate?

Very roughly. AFAIK, those numbers are 'scraped' from old reviews, etc. Esp. with Intel, there have been massive increases to performance since the sourcedata was published. (Also, 'AMDfinewine' and a few cases of nVidia actively gimping older cards, via Drivers)

It's not looking good to me, performance is similar to a 7900 XT and the same is true in Linux, too.

Considering the pricing and availability of the XT and GRE 7900s... Navi48 is still 'better'.

Navi48 presents at $550-700, for $550-800+ of today's performance.
-in a slightly smaller, slightly cooler, slightly more efficient package, with notably better RT performance to boot.

You're not entirely wrong, though.
If nVidia hadn't been the trendsetter they have been, this AMD launch would be absolutely underwhelming.

At this point though, all AMD has to do to 'be amazing' is launch a card that
doesn't melt,
comes with all the hardware on the specsheet,
and is actually available for sale (in qty).

Prices are probably going to skyrocket after current stock sells out as well.

Looking back at the past couple-few years, and what pricing/availability looks like today, that seems obvious.

Am I the crazy one?

Nope. Maybe a tinge pessimistic.

TBQH, though...
I'd say a good lot of us are a tad 'demoralized', by the way the market's been.

 

[alwayssts]

Are the relative performance percentages in the GPU database accurate? It's not looking good to me, performance is similar to a 7900 XT and the same is true in Linux, too. Prices are probably going to skyrocket after current stock sells out as well. Am I the crazy one?

Never, in the history of ever in videocards, has there been a worse time to judge everything on one comparative percentage graph. Let's just go with that and then we can all agree.

 

[Jism]

Great card, great review. TPU always on top of their game.

 

[Rover4444]

Very roughly. AFAIK, those numbers are 'scraped' from old reviews, etc. Esp. with Intel, there have been massive increases to performance since the sourcedata was published. (Also, 'AMDfinewine' and a few cases of nVidia actively gimping older cards, via Drivers)

Considering the pricing and availability of the XT and GRE 7900s... Navi48 is still 'better'.

Navi48 presents at $550-700, for $550-800+ of today's performance.
-in a slightly smaller, slightly cooler, slightly more efficient package, with notably better RT performance to boot.

You're not entirely wrong, though.
If nVidia hadn't been the trendsetter they have been, this AMD launch would be absolutely underwhelming.

At this point though, all AMD has to do to 'be amazing' is launch a card that
doesn't melt,
comes with all the hardware on the specsheet,
and is actually available for sale (in qty).

Looking back at the past couple-few years, and what pricing/availability looks like today, that seems obvious.

Nope. Maybe a tinge pessimistic.

TBQH, though...
I'd say a good lot of us are a tad 'demoralized', by the way the market's been.

Pessimistic yeah, but the 7900 XT has been under $700 for nearly a year. It might have better RT, but it also has 4GB less VRAM so I think the value tapers out a bit in that regard...

36% better performance at a 20% higher price means only a 13% better value over the 7800 XT at it's MSRP. That's way worse than I was predicting, I really think this card should have been $550.

Never, in the history of ever in videocards, has there been a worse time to judge everything on one comparative percentage graph. Let's just go with that and then we can all agree.

It's a good thing I'm using two of em! Phoronix is saying the same thing for Linux performance. Maybe things will get better over time, but they've already delayed this launch for three months. I'd think it'd be a bit better upon release, right?

 

[-1nf1n1ty-]

Am I crazy for wanting to sell my 4080 super to get the 9070XT and pocket the extra money?

 

[mkppo]

This is a false equivalence. Cyberpunk 2077 and Resident Evil 4 are acclaimed, successful, and well-optimized games that fully leverage the potential of each architecture. Silent Hill, on the other hand, is a game that simply runs poorly on everything; the computational load compared to the visuals on screen is absurd.

To be honest, I wouldn't include in a GPU review any game where high-end GPUs struggle at 30–40 FPS or where performance is excessively high, like 300–500 FPS.
It would be nice if, in the future, we could at least vote to include games we like.

Glad I'm not the only one who has qualms with SH2 and CS. The latter runs at an ungodly FPS anyway and SH2 is just plain crap when it comes to optimization and looks like shit to boot.

I would love to see some good console ports though, like Forza motorsport 7 or even the old ish FH5.