AMD Radeon RX 7900 XTX

[Cheeseball]

They need to get their drivers to read the monitor's EDID properly so that the idle clocks would be on par with both NVIDIA and Intel. Using CRU is still the workaround for that issue.

After that, pretty much Adrenalin is fine for what it is (aside from Enhanced Sync being useless nowadays).

Yes, of course I've reported this issue through their Bug Report tool. It's been 3 years actually and still no notable movement on this. (Find my 5700 XT thread)

 

[ChaoticG8R]

Im optimistic that the board partners will push the XTX even farther. It’s fairly obvious that nvidia has put some incredible effort to ensure their internal cooler/design is competitive with AIBs, while it feels more like a true bare minimum/“reference” style that AMD is providing.

I would assume we’ll have some reviews tomorrow on launch for the AIBs @W1zzard or is AMD going to stagger them again like previously? If you can’t give a yes or no due to NDA, can you at least acknowledge that?

Thanks in advance

 

[TheoneandonlyMrK]

I'm gonna leave this here
Top Nvidia shareholders

Vanguard Group Inc. representing 7.7% of total shares outstanding​

BlackRock Inc. representing 7.2% of total shares outstanding​

Top AMD shareholders

Vanguard Group Inc. representing 8.28% of total shares outstanding​

BlackRock Inc. representing 7.21% of total shares outstanding​

Brilliant post, again, , ,,Not.
I think the thread your after is the actual project Calisto poor performance thread.
You can shitpost in peace there.

OP equals a 3090Ti in raytacing, edges the 4080 in 4k raster and cheaper than the competition, not bad, not the outright win many(I) wanted but not bad.

Come on Santa you Shit, sort it out.(unlikely but I can dream)

 

[Ravenas]

Performance/watt should be larger on AIB cards given 3 and 4 8 pin connectors. You will see larger power draw with larger performance increases.

Energy efficiency. Clear advantage over 4080 in regard to rasterization and price. Ray Tracing in 4K still isn’t ready for prime time yet.

@W1zzard The argument that you should by a 4080 over a 7900 XTX for ray tracing is ludicrous in my opinion. Averaging 4K 29 FPS on 4080 versus 21 FPS 7900XTX in Cyberpunk is comical. It’s almost across the board game to game.

Well deserved Editors Choice.

 

[Luminescent]

I have a fan.

 

[kapone32]

Sorry, i'll be keeping my 6800XT for another gen i guess.

That is fine it is your money to do with as you please. I play TWWH3 and when I saw 145 FPS in the review I was sold.

 

[N3M3515]

barely beating the 4080.

It's identical in raster, unless you can detect a 3% difference?
RT nvidia +16%.

But both are $700 gpus.

I saw 145 FPS in the review I was sold.

Yeah, that's sad, it's like approving those price hikes. Next gen 8900XTX $1500. And you'll see 185 FPS in TWWH3 and be sold again.

 

[Bjorn_Of_Iceland]

If I was in the market for a GPU, I'd rather go for a used RTX3090 than a 7900 XTX tbh.

 

[Alpha_Lyrae]

Ah, yes, launch driver bugs; very typical of AMD. You know, more of you should be demanding “FineWine” performance at launch and not 6-12 months later. That simply means that the driver compiler is still not optimized for RDNA3 after all of this development time, and yes, it’s resource intensive, but most people base decisions on launch performance.

Overall, RDNA3 isn’t terrible, however we should all remember that this generation, AMD and Nvidia are both on TSMC again. Nvidia is using “4N” or Nvidia-optimized N4(P?) while AMD is using N5 (also with AMD customized libraries) for GPU and N6 for MCDs.

They’re not quite on equal terms: N4P (closest to 4N) offers 11% more performance and 22% more efficiency over N5 (just 4% and 7% over N5P). Density is only a 6% reduction vs N5. These are TSMC’s own numbers. AMD can refresh Navi 31 GPU in 8-12 months with N4P at higher wafer costs and that would close pricing between AMD 7950XTX and eventual Nvidia RTX 4080Ti.

Nvidia screwed themselves with Ampere by going to Samsung 8LPP.

 

[Vayra86]

Compact and loud go together. Too bad, why can't they design these things better?
But other than RT performance it's quite great.

The XT is only about 20w more than 6800xt, that's a high increase in efficiency

39dB at 58C core temp is pretty neat I'd say, you can't look at noise in isolation imho.

These either OC to the moon or they can go pretty damn silent under load, being as they are in the review; on the OC page you can see the XTX hitting near 3 GHz.

It does not however extract a whole lot of FPS from it in Heaven, and not all of that I think is attributable to the CPU limitations. The lack of refinement we see there echoes throughout the review; apparently it clocks very aggressively at the beginning of a run and is quickly forced to push back to remain within 350W power target.

That is fine it is your money to do with as you please. I play TWWH3 and when I saw 145 FPS in the review I was sold.

Yeah somehow that game cripples my 1080 too, I'm looking at 30 FPS on the campaign map. Still wondering why it drops so low; TWWH2 does 50% more on average. Its not exactly paid off by massive increase in fidelity...

It's identical in raster, unless you can detect a 3% difference?
RT nvidia +16%.

The RT results AND raster results are pretty wide apart in some cases, but if you consider that and then still consider there's 3% in advantage of AMD, that's meaningful. Some games push 20% more FPS, while the biggest loser is what, 15% in favor of Nvidia. The number of games is also higher where it scores better. Given the lacking optimization elsewhere in the product, its safe to say that gap can increase further in favor of AMD. Not a given, but definitely plausible.

 

[Makaveli]

how about PRODUCTIVITY TESTS?
x264, AV1 H265
CONDING ENGOCING DECODING VIRGINING?

 

[AnotherReader]

It's identical in raster, unless you can detect a 3% difference?

We agree; you didn't quote the full sentence.

I was expecting the 7900 XTX to be around 90% of a 4090 in traditionally rasterized games, but it is barely beating the 4080

 

[kapone32]

Yeah, that's sad, it's like approving those price hikes. Next gen 8900XTX $1500. And you'll see 185 FPS in TWWH3 and be sold again.

You can call it what you want. Who knows, it depends but yes I do buy AMD cards. Even the Vega 64 has been worth every single penny and my RX570 is still going strong too. I am not taking food out of my mouth to buy a $1000 GPU so it really is inert. I do not like the prices but a PS5 is $999 all day long and pre-builts with cards like the 3090 and TI versions are still $4000+ so it is what it is.

 

[medi01]

• Considerably lower ray tracing performance than RTX 4080

Cough.

Ray Tracing in 4K still isn’t ready for prime time yet.

No offense, 3090Ti.

 

[Sir Alex Ice]

So what happens if I connect a Dell U2720Q or similar monitor that has a Type-C display port using the Type-C DisplayPort 2.1 in the 7900 XTX?

 

[Vayra86]

• Considerably lower ray tracing performance than RTX 4080

Cough.


View attachment 274193


No offense, 3090Ti.

Yeah the RT perf should not be a deal breaker on its own, I do agree.

But then you see that the perf/$ gap is also exactly 17% to the 4080, and it kinda does get ya thinking. That balance isn't tipping decisively in AMD's favor there; after all, RDNA3 is missing the mark on a few other aspects, most notably idle power plus a historical drawback; people are seeing history repeat: cards getting released that could have been refined more.

And to be fair to them, I see that too. New technology, blah blah fantastic, but if that serves to brute force your way to the sub top of leaderboards while leaving refinement on the table, that ain't good. RDNA2 was great exactly because it did tick all those boxes. And that's a literal repeat of what we saw since Fury. HBM never materialized in the gaming space. If MCM isn't paying off bigtime, what's it doing here? After all, if it offers cheaper products for being non monolithic, why aren't we seeing it?

I'm kinda withholding judgment because I expect two things in the near future;
- price drops
- driver refinement

 

[75Vette]

Greetings. Awesome review as always... small error though:

4090 is $2400 now???

This is in all the charts on that page: AMD Radeon RX 7900 XTX Review - Disrupting the RTX 4080 - Performance per Dollar | TechPowerUp

 

[Cheeseball]

So what happens if I connect a Dell U2720Q or similar monitor that has a Type-C display port using the Type-C DisplayPort 2.1 in the 7900 XTX?

It should just consider it a normal DisplayPort connection.

 

[Denver]

It's identical in raster, unless you can detect a 3% difference?
RT nvidia +16%.

But both are $700 gpus.


Yeah, that's sad, it's like approving those price hikes. Next gen 8900XTX $1500. And you'll see 185 FPS in TWWH3 and be sold again.

Why? Did you roll a dice and decide on that price? The reality is that contrary to what you think these GPUs don't have very high profit margins like in the past, $700 should be more or less the cost of production. Over $300 just for memory (optimistic), it would be great if you could maintain that performance with just 12Gb or 16gb of vram, so it could cost $150-200 less

My main and only complaint is in relation to the huge variation in performance difference between games, in some cases it even beats the 4090 and in others it loses by about 20% to the 4080. Plus, It seems a rule that games based on the Unreal Engine run bad AMD gpus, I wonder if AMD doesn't have money left over to do something about it.

Anyway, it's still the most interesting product AMD has released in recent years, too bad there aren't decent games to effectively use so much computational power.

 

[Dristun]

Anyway, it's still the most interesting product AMD has released in recent years, too bad there aren't decent games to effectively use so much computational power.

IMO 6900XT was way more interesting - actually faster than nvidia's best in raster, and a time when ray-tracing was less relevant and in way fewer games. This gen feels like a geniune step back again - they can't even comprehensibly beat a cut-down die while still lagging in features (or gimmicks, if you wish, it doesn't really matter because one has them and the other doesn't, at least right here and right now without FSR3 and so on).

 

[Vayra86]

Why? Did you roll a dice and decide on that price? The reality is that contrary to what you think these GPUs don't have very high profit margins like in the past, $700 should be more or less the cost of production. Over $300 just for memory (optimistic), it would be great if you could maintain that performance with just 12Gb or 16gb of vram, so it could cost $150-200 less

My main and only complaint is in relation to the huge variation in performance difference between games, in some cases it even beats the 4090 and in others it loses by about 20% to the 4080. Plus, It seems a rule that games based on the Unreal Engine run bad AMD gpus, I wonder if AMD doesn't have money left over to do something about it.

Anyway, it's still the most interesting product AMD has released in recent years, too bad there aren't decent games to effectively use so much computational power.

Historically Nvidia has ran margins of 40-60% and they're peaking above 60% today. So strangely enough as nodes shrink, both revenue and margins shoot up, and not one at cost of the other.

700 cost of production? Got source, or just pulling it out of what's been going around the internet in random blurbs? I'm interested if you have real info. And if AMD has to cut into its margins so much more than Nvidia doing its monolithic approach, what the fuck is AMD even on right now? Crack? Or is this the only way they can realistically move forward? Or is it an investment in the future? If so, they should damn well make sure to release it in a better state.

The story just doesn't make sense. If MCM is better wrt yields on smaller nodes, it should end up cheaper.

Note; here's AMD's gross margin. Note that CPU is chasing the chiplet route for a while now too.

 

[john_]

I'm a huge fan of new intel, but they're not even close to 'biting' AMD from behind. If intel can match the 6800xt with battlemage next year I will be impressed, but I doubt it.

This architecture does have similarities to bulldozer, but at the end of the day it pushes the FPS needed to compete with the 80 class.

They will need to drop price, for sure -- but hopefully the silicon savings allow for that.

They don't need to reach 6800X level of performance. Nvidia is securing the high end and Intel will take market share from AMD in the low-mid range market by selling to big OEMs. Intel can sell multiple times more products than AMD to OEMs, even inferior products. Intel is also doing something smart. It's investing from the beginning in RT performance. AMD is the only one of the three GPU manufacturers that still considers RT performance secondary. That's a mistake in my opinion. And now that they are having problems destroying RTX 4080 in raster, it's going to cost them heavily in sales. I can already see a drop of $100 in price, at least for the XT, even if Nvidia doesn't drop RTX 4080's price, or/and RTX 4070 Ti comes at $800. I agree with you in this one.

 

[Fluffmeister]

IMO 6900XT was way more interesting - actually faster than nvidia's best in raster, and a time when ray-tracing was less relevant and in way fewer games. This gen feels like a geniune step back again - they can't even comprehensibly beat a cut-down die while still lagging in features (or gimmicks, if you wish, it doesn't really matter because one has them and the other doesn't, at least right here and right now without FSR3 and so on).

Yeah that's a good point, the 4080 isn't using a fully enabled AD103, and then worringly neither is the 4090 close to fully using the AD102. A fully enabled AD102 on presumably the 4090 Ti is gonna bitch slap everything (and melt wallets).

 

[W1zzard]

0.1 V is too low, you need at least 0.5-0.6 V to open transistor. Clocks aren't real too.

This is what the monitoring circuitry in the card returns, I have no way to disprove these results

View attachment 274179View attachment 274182

TFW The 7900XTX pushed balls to the wall beats top end RDNA2 by a whopping ~43FPS in raster.
View attachment 274181

5800X vs 13900K though. Check the deltas for cards present in both tests

These either OC to the moon

Actually these don't OC at all, as mentioned in the review. When you change GPU clock you get 3 results:
- No change in performance
- Loss in performance
- Crash

What you don't get is any performance gain.

What works for "OC" is undervolt, so the boosting increases clocks (without you manually touching clocks), and increasing power limit, so the boosting increases clocks (without you manually touching clocks).

Others would say "OC is broken", if that's the way it works now.

Greetings. Awesome review as always... small error though:

4090 is $2400 now???

That's the current market price. I tried to find one at lower price this morning, not possible, all sold out.

 

[Vayra86]

Actually these don't OC at all, as mentioned in the review. When you change GPU clock you get 3 results:
- No change in performance
- Loss in performance
- Crash

What you don't get is any performance gain.

What works for "OC" is undervolt, so the boosting increases clocks (without you manually touching clocks), and increasing power limit, so the boosting increases clocks (without you manually touching clocks).

Others would say "OC is broken", if that's the way it works now.

Yeah I noticed that, and the GPU is also at peak power all the time. Looks remarkably similar to Zen4.

What do you estimate the chances of this changing with future updates?