9070 and 9070XT power consumption leaked by HUB

[Cowboystrekk]

Hmm...Nice.

https://x.com/GawroskiT/status/1896838352844026103

That is impressive

 

[Onasi]

That is impressive

You’d think so, but RDNA 2 was notoriously shit in RT. Add a decent overall performance uplift and this is less “impressive” and more “actually what it should be to compete”.
It should also be noted that we are talking what is essentially 1080p with frame gen. Not surprising for heavy RT, but still kinda stings when put in perspective. Granted, NV ain’t doing much better here once you turn up their gimmicks up on mid-range cards either.

 

[alwayssts]

That is impressive

I would say so, yes.

As always, I look at mins. But still, if 1080p->4k raytracing is actually realistic on this thing...That's pretty sweet.

Ofc, I wonder how it (FSR4) looks. I'm sure it isn't bad, and will only get better, but I'm not going to go bananas just yet given how that ToyShop demo looked.

I'll say it again, I do champion good framerates across resolutions (and up-scaled resolutions) with this technology over better IQ that is not, but, you know, better IQ is better than worse IQ.

 

[BoggledBeagle]

Apparently this results are not gpu only, so it would be better to judge by differences rather than absolute values while comparing. Eg. 9070XT system consumes about 80w more than 5070TI.

It does not look good at all. It seems that they clocked the 9070XT as far as it was possible out of the box.

I wonder how idle power draw will look like, my MSI 5070 Ti somehow refuses to draw less than 30W with two monitors hooked up.

 

[mb194dc]

So the 9070xt likely is similar to a 4080s in performance? That's how I would read those charts.

 

[alwayssts]

So the 9070xt likely is similar to a 4080s in performance? That's how I would read those charts.

Accurate, afaict. This is my interpretation right now (we'll know more soon):

If similar-ish to 4080:

Stock: slightly under 60fps 1080p BMW
OC: 1080p60mins in BMW
OC + 4k up-scale + FG: 60fps perhaps in BMW (with perhaps tolerable latency)

If that's true, and it's really important if it is or isn't IMHO, that's a win. That's exactly what people need. I hope IQ is good, I hope it gets better, but that's a good 'mid-range' card RN imho if it can do that.

I'm also very curious how it will do in Outlaws at 1440p and 1440p->4k. These are kind of important dividing lines (imho), and *could* be what separates something like a XT/XTX (or future tiers of cards).

If it can all of those things okay, I'll be impressed.

 

[Caring1]

Doesn't look like they improved efficiency much (based on performance claims) and the cards are still very power hungry compared to Nvidia equivalent.
Apparently this results are not gpu only, so it would be better to judge by differences rather than absolute values while comparing. Eg. 9070XT system consumes about 80w more than 5070TI.

Tell me you're an Nvidia fanboi without telling me. *sigh.

 

[alwayssts]

Tell me you're an Nvidia fanboi without telling me. *sigh.

I mean, it's fair to an extent, though, right?

I mean, I very much am on the side of cheaper cards with less units that can (over)clock higher, even if they use more power.
Some are in the camp of lower-clocked cards with more units that are more energy-efficient. (That you'll desire to upgrade every gen bc planned obsolescence.)

Both are fair opinions depending upon how you view the world. I will always argue people are not stressing their card 100% 24/7 etc, and to think about that, but it truly is somebody's prerogative.

I just ask people keep an open mind and be realistic; it's fine if you value one thing over another (especially, for instance, if you're running a SFF with limited thermal allowance, etc).
If you like to tinker and play around, sometimes power be damned, hoping that the limit of your card will surpass the next setting you can toggle at 60fps mins, which AMD card's often can, then that's cool too.

In many ways it really just depends on what kind of person you are, and what you're trying to accomplish.

 

[Caring1]

I mean, it's fair to an extent, though, right?

I mean, I very much am on the side of cheaper cards with less units that can (over)clock higher and cost less money, even if they use more power.
Some are in the camp of lower-clocked cards with more units that are more energy-efficient. (That you'll desire to upgrade every gen bc planned obsolescence.)

Both are fair opinions depending upon how you view the world. I will always argue people are not stressing their card 100% 24/7 etc, and to think about that, but it truly is somebody's prerogative.

I just ask people keep an open mind and be realistic; it's fine if you value one thing over another (especially, for instance, if you were running a SFF with limited thermal allowance, etc).

I want AMD and Nvidia to get their act together and lower power consumption without loss of performance. Every other industry can do it, why can't they.

 

[Dr. Dro]

Hmm...Nice.

https://x.com/GawroskiT/status/1896838352844026103

Frame Generation seems to be enabled, and 19 fps seems like a notoriously poor result, especially with frame generation enabled. This would be a supreme edge case, mostly because it exposes the absolute worst case scenario for a RX 6800 XT - which was horrible at ray tracing to begin with. Personally, I expect somewhere around 80% faster than 6800 XT in "real world" RT workloads, less in pure raster, which should slot it in the RTX 4070 Ti Super's performance level or thereabouts. Guess we'll see in a couple of hours, although I'm probably going to be asleep when W1zz publishes the reviews.

Tell me you're an Nvidia fanboi without telling me. *sigh.

Given the leaked results, it does seem like the extra 100 to 120 watts over the 5070 non-Ti is relatively consistent, although it should also be faster. I did notice something though, the 5070 uses a good chunk more power than the 4070 did. GDDR7 is supposed to be more efficient, so I'll chalk that to more aggressive clock speeds and more shaders, I suppose.

 

[alwayssts]

I want AMD and Nvidia to get their act together and lower power consumption without loss of performance. Every other industry can do it, why can't they.

Many times I feel it comes down to optimal configurations (unit/clock ratios that line up for the ability of libraries on a process matched with available bw/buffer options), and then somewhat matching markets. Sometimes that means optimal perf/w, sometimes maximum perf per design. It all depends.

I really do think it could be potentially interesting next-gen, if say AMD continues to use a high-clocking process and nVIDIA a more power-efficient one, with similar designs...what will people choose?

It would be then people would have to put their money where their mouth is.

Frame Generation seems to be enabled, and 19 fps seems like a notoriously poor result, especially with frame generation enabled. This would be a supreme edge case, mostly because it exposes the absolute worst case scenario for a RX 6800 XT - which was horrible at ray tracing to begin with. Personally, I expect somewhere around 80% faster than 6800 XT in "real world" RT workloads, less in pure raster, which should slot it in the RTX 4070 Ti Super's performance level or thereabouts. Guess we'll see in a couple of hours, although I'm probably going to be asleep when W1zz publishes the reviews.



Given the leaked results, it does seem like the extra 100 to 120 watts over the 5070 non-Ti is relatively consistent, although it should also be faster. I did notice something though, the 5070 uses a good chunk more power than the 4070 did. GDDR7 is supposed to be more efficient, so I'll chalk that to more aggressive clock speeds and more shaders, I suppose.

Right, but it's not about the 6800xt at all, which is (pretty much) a raster card. It's about what the 9070 xt can do.

Given your expectations, I think you will be pleasantly surprised. I mean, again, I think it's going to be like a 4080 more-or-less when it's OC'ed. Is it miraculous? No. Good value? Yeah.

 

[Dr. Dro]

Cautiously optimistic, really. I'm hoping to be pleasantly surprised when I wake up. They might just be able to steer the ship around, the press releases of a few days ago left me satisfied.

The lead time on my RTX 5090 was meant to be around ~68 days (currently we're at day 35 since launch day event), and that's more than I'm honestly willing to wait as it is. If there are any further delays I think I'll cancel my order and buy one of these.

 

[Caring1]

Jayz2Cent gave a sly hint they will be worth waiting for

 

[Assimilator]

I want AMD and Nvidia to get their act together and lower power consumption without loss of performance. Every other industry can do it, why can't they.

Because you don't understand the fundamental physical limitations of silicon fabrication at the Moore's Law wall. Go away and learn that and then you won't post such stupid things.

 

[droopyRO]

I bet that undervolting does miracles once again.

I am a strong advocate for undervolting. But i did not have any luck with my 6650 XT, unless i drop the frequency too. On the 7xxx and upcoming 9xxx there might be a better results.

 

[alwayssts]

I am a strong advocate for undervolting. But i did not have any luck with my 6650 XT, unless i drop the frequency too. On the 7xxx and upcoming 9xxx there might be a better results.

Well, in what regard? In the respect they may be binned at a higher voltage (for yield) and then power-limited, so then you under-volt them to see where it will hit the PL while clocking higher?

I mean, I don't really call that under-volting. I call that the new overclocking. If you're *really* under-volting, you're trying to find the point of diminishing returns.
In that respect, I think these will be well-matched at stock. If you're not going to touch the memory, you might be able to undervolt ~10% and lower PL ~20%, but you could just as well up everything ~10%.

In the end, you might save 20% power or gain 10% performance for 20% more power, but in the grand scheme of these cards I think it's going to be worth pushing them as they *appear* to be on the brink of a lot of settings at 60fps. JMO based on what we've seen. We'll have to see which side of the fence they typically fall at stock, and then people can/will correct accordingly.

 

[mama]

View attachment 387817

These numbers are quite odd. Even so, it will still be more efficient than the previous generation, and the 9070 will undoubtedly be more efficient than the 5070.

Someone has been busy with photoshop.

 

[Darkholm]

Looking good

 

[BoggledBeagle]

View attachment 387907

Looking good

Post more!

Oh, you just copied a picture from Videocardz without giving them credit, do not bother then.

Last-minute Radeon RX 9070 performance leaks appear before embargo - VideoCardz.com

Radeon RX 9070 in last-minute leaks First up is the 3DMark Time Spy Score featuring both Radeon RX 9070 cards. We can confirm that these are indeed real scores for these cards, where the Radeon RX 9070 XT is set to score around 30,050 points in the Graphics test, whereas the non-XT version...

videocardz.com

 

[3x0]

Oh, you just copied a picture from Videocardz without giving them credit, do not bother then.

It's from TechYesCity on YouTube

 

[Darkholm]

Post more!

Oh, you just copied a picture from Videocardz without giving them credit, do not bother then.

Last-minute Radeon RX 9070 performance leaks appear before embargo - VideoCardz.com

Radeon RX 9070 in last-minute leaks First up is the 3DMark Time Spy Score featuring both Radeon RX 9070 cards. We can confirm that these are indeed real scores for these cards, where the Radeon RX 9070 XT is set to score around 30,050 points in the Graphics test, whereas the non-XT version...

videocardz.com

It is YouTube screenshot from TYC and their logo is visible.

 

[BoggledBeagle]

I knew the source, you just made me search for more without a chance for finding more.

 

[Darkholm]

I knew the source, you just made me search for more without a chance for finding more.

Sorry about that, I was posting via mobile phone while taking a sh1t.

 

[stahlhart]

"Puke"?

 

[Scattergrunt]

I want AMD and Nvidia to get their act together and lower power consumption without loss of performance. Every other industry can do it, why can't they.

Yea, we will eventually start hitting i9-9900k levels of insanity with the TDP's on GPU's, and then something will kick them in the balls to make them go 'okay maybe we shouldn't just turn up the TDP..", for Intel, it was the higher TDP 13th and 14th gen processors going caput.