News AMD Radeon RX 7900 GRE partner card roundup — all the cards currently listed

Status
Not open for further replies.
Looking at the performance reviews from Hardware Unboxed and Gamers Nexus, this card's existence is completely pointless. It's $50 more expensive than the RX 7800 XT for what amounts to pretty much the same performance.

I honestly don't know why AMD bothered.
 
It's another cheaper card with 16GB AMD at least has that going for them, how many 16GB GPU's does AMD have for sale compared to Nvidia right now?
 
Looking at the performance reviews from Hardware Unboxed and Gamers Nexus, this card's existence is completely pointless. It's $50 more expensive than the RX 7800 XT for what amounts to pretty much the same performance.

I honestly don't know why AMD bothered.
Because it's 5~10 percent faster, and that makes it look a bit better in the benchmarks? RTX 4070 was matching or beating the 7800 XT, so adding a bit more performance makes things look better.

Also because there are probably enough Navi 31 GCDs that either don't have enough functional CUs (maybe), or with more than one defective MCD link (also maybe), or perhaps with some other defect that prevents them from being sold as a 7900 XT.
 
t
Because it's 5~10 percent faster, and that makes it look a bit better in the benchmarks? RTX 4070 was matching or beating the 7800 XT, so adding a bit more performance makes things look better.

Also because there are probably enough Navi 31 GCDs that either don't have enough functional CUs (maybe), or with more than one defective MCD link (also maybe), or perhaps with some other defect that prevents them from being sold as a 7900 XT.
aside from gaming there would surely be an nice uplift in stable diffusion ONNX or AI related work loads because it has way more shaders?
 
aside from gaming there would surely be an nice uplift in stable diffusion ONNX or AI related work loads because it has way more shaders?
Stable Diffusion (on AMD at least) ended up being pretty seriously bandwidth limited in my testing.

5HhxX2zYeJUMPh7Zdfhog8.png


8QKTDhr4Mkz7dSdDrKreQ8.png
 
Because it's 5~10 percent faster, and that makes it look a bit better in the benchmarks? RTX 4070 was matching or beating the 7800 XT, so adding a bit more performance makes things look better.
Yeah, I suppose. It's just that it's also 10% more expensive so that makes it pretty underwhelming overall. If it were just 15% faster than it is, it would be more at a mid-point between the RX 7800 XT and RX 7900 XT which I believe would make it considerably more exciting than it is. A lot of the numbers that I saw from Hardware Unboxed were single-digit FPS increases (some being VERY LOW single-digit increases) and that was pretty disappointing.
I honestly expected that the RX 7900 GRE would've been at least 15% faster than the RX 7800 XT (putting it at that magic midpoint between the 7800 XT and 7900 XT) but, to be fair, since the GRE came out so long ago and the 7800 XT came out so recently, I had long forgotten the performance numbers of the GRE.
Also because there are probably enough Navi 31 GCDs that either don't have enough functional CUs (maybe), or with more than one defective MCD link (also maybe), or perhaps with some other defect that prevents them from being sold as a 7900 XT.
I get that, I'm just not sure that AMD is doing enough to make these cards palatable to consumers. Paying 10% more for 5-10% more performance would probably strike most people as just not worth it because it doesn't take any games that are unplayable on the 7800 XT and make them playable. The difference between them would be imperceptible, except the extra $50 missing from the wallet. 😉
 
Stable Diffusion (on AMD at least) ended up being pretty seriously bandwidth limited in my testing.

5HhxX2zYeJUMPh7Zdfhog8.png


8QKTDhr4Mkz7dSdDrKreQ8.png
oh sorry how was stable diffusion configured? or what version was just stable diffusion ML? or was it run using the ONNX optimized version?

I tested 2 versions one was stable diffusion ML but the other was the master version on GIT HUB that version allows you to optimize models with ONNX within the webui and with that version my old amd card more than doubled it speed compared to the normal WebuiML that pull's and installs ONNX as part of the install package but didnt have the options to optimize models within the webui

my installs were done in CMD using git hub to pull the ML version and the master version was the version that is a zipped file on github basicly the zipped version here

https://github.com/lshqqytiger/stable-diffusion-webui-directml

For it to work you have to Install and use an older version of python, python version 3.10.6 beacuse something to do with pytorch i think cant remember now it been a little while since I messed around with it, but if you have a newer version of python set up and configured it probs not worth messing around but I dont know I'm just a noob. you probs know way more than me about it.
 
Last edited:
oh sorry how was stable diffusion configured? or what version was just stable diffusion ML? or was it run using the ONNX optimized version?
Yes, ishqqytiger is the version I used on AMD. You have to "recompile" or whatever the ONNX model for each different GPU. Lots of fun! But you have to do something similar for Nvidia's latest TensorRT stuff.

Yeah, I suppose. It's just that it's also 10% more expensive so that makes it pretty underwhelming overall.
See that's the problem, though: 10% more money for 10% more performance is actually a good deal overall. I mean, look at how many parts are 50% more money for 10% more performance, or at least less than 20%. And it's not just GPUs, it's CPUs, SSDs, memory, etc. I need to figure out a good way to generate a table or chart for this, as I have all the data in my spreadsheets... Here we go (not based on MSRP, though):

GPURetail PricePerf/$Value Rank
RTX 4090$1,9000.088943
RTX 4080 Super$1,0500.141836
RTX 4080$1,1860.122639
RTX 4070 Ti Super$8000.167431
RX 7900 XTX$9200.142135
RTX 4070 Ti$7200.177028
RTX 3090 Ti$1,7260.072844
RTX 4070 Super$5900.202712
RX 7900 XT$7000.170930
RTX 3090$1,2800.092442
RTX 3080 Ti$1,0290.112841
RTX 3080 12GB$8950.126937
RTX 4070$5250.201913
RX 7900 GRE$5500.192521
RTX 3080$5750.182924
RX 6950 XT$5800.177727
RX 6900 XT$7800.126038
RX 7800 XT$4900.200115
RX 6800 XT$4600.201714
RTX 3070 Ti$5000.174829
RX 7700 XT$4200.206210
RTX 4060 Ti 16GB$4300.193120
RTX 4060 Ti$3750.22065
RTX 3070$4250.193819
RX 6800$4010.203411
RTX 3060 Ti$4650.158333
RX 6750 XT$3600.195917
RTX 4060$2850.23623
RX 6700 XT$3500.189523
RX 7600 XT$3200.191722
Arc A770 16GB$2900.20919
RTX 2070 Super$5000.118840
RX 6700 10GB$3000.193818
Arc A770 8GB$3160.179325
Arc A750$2100.26432
RTX 3060$2800.198616
RX 7600$2600.21337
RX 6650 XT$2410.21956
RX 6600 XT$2400.21198
Arc A580$1650.30461
RX 6600$1900.22544
RTX 2060$2790.152534
RTX 3050$2250.178026
Arc A380$1150.165332
 

And that exemplifies the issue surrounding AMD's pricing. You can get an ASUS Dual GeForce RTX 4070 SUPER for the same price, and the GIGABYTE GeForce RTX 4070 SUPER WINDFORCE OC, and the ZOTAC GAMING GeForce RTX 4070 SUPER Twin Edge (actually $10 less right now on "sale").

With the 4070 Super performing effectively (1-3% difference overall) the same in games while performing 25% better when ray tracing is involved, and having far better efficency (TechPowerUp arrived at the same power figures in their tests as well), outside of people who need to use the programs AMD specifically does better in than nVidia, how do you tell someone to get the 7900 GRE over the 4070 Super?

Qdgk3MfBr4HGrKRAMbwRKa.png
 
Yeah, I suppose. It's just that it's also 10% more expensive so that makes it pretty underwhelming overall. If it were just 15% faster than it is, it would be more at a mid-point between the RX 7800 XT and RX 7900 XT which I believe would make it considerably more exciting than it is. A lot of the numbers that I saw from Hardware Unboxed were single-digit FPS increases (some being VERY LOW single-digit increases) and that was pretty disappointing.
I honestly expected that the RX 7900 GRE would've been at least 15% faster than the RX 7800 XT (putting it at that magic midpoint between the 7800 XT and 7900 XT) but, to be fair, since the GRE came out so long ago and the 7800 XT came out so recently, I had long forgotten the performance numbers of the GRE.

I get that, I'm just not sure that AMD is doing enough to make these cards palatable to consumers. Paying 10% more for 5-10% more performance would probably strike most people as just not worth it because it doesn't take any games that are unplayable on the 7800 XT and make them playable. The difference between them would be imperceptible, except the extra $50 missing from the wallet. 😉
However, HwU only tested a stock card. And none of the new batch of cards are at stock. They have base speeds up to 394 MHz (31%) faster and boost clocks up to 146 MHz (6.5%) faster. Techpowerup tested one of the faster OCed cards and found it was 4% more powerful overall than the stock card.

While those aren't stellar figures, they're better than HwU showed, and explains why almost every other reviewer/tester had better results for the 7900 GRE. And it also helps to justify the extra cost over the 7800 XT, pushing it closer to the 10% extra performance to go with the 10% extra cost.

While it's true that some of the OCed cards are $10-50 more than the MSRP, not all of them are. I have the Gigabyte Gamer OC one on the way, and it's the same $550 price as stock, but with those OCed speeds.

Also, remember that there are lots of different potential customers with different needs and expectations and acceptable price ranges. What doesn't make sense for you definitely makes sense for me. And probably for a lot more people than you realize. And if not, then AMD will have to drop the price, which is fine.

For me, it would have been hard to justify upgrading my 6800 XT to the 7800 XT (which is why I didn't!) with the very minimal performance uplift. But the 7900 GRE hits that sweet spot with actual, noticeable and measurable improvement, but without the high cost of the 7900 XT, which I also couldn't justify. Remember that while it's not at the midpoint between the 7800 XT and the 7900 XT in performance, it's also not at the mid-point in price either.

And it uses less power than my 6800 XT or the 6900 XT, so I don't have to upgrade my PSU or worry about power issues. And I can always sell off my 6800 XT to help justify it, making the upgrade around $200, and allowing me to start to move into 4k gaming.
 
  • Like
Reactions: Avro Arrow
See that's the problem, though: 10% more money for 10% more performance is actually a good deal overall.
Don't get me wrong I know it's a good deal. Everybody knows that value decreases as you go up the product stack. I didn't mean to say that it wasn't a good deal, it's just nothing new and exciting. It's kind of like how, sure the R7-5700X3D is a great deal, but there was no mystery to it at that point. I also was hampered by the fact that I had only seen Steve Walton's video and I didn't know until today that he did it wrong. Now that I've seen videos from Steve Burke and Daniel Owen, I realise that the card is in fact not pointless.
However, HwU only tested a stock card. And none of the new batch of cards are at stock. They have base speeds up to 394 MHz (31%) faster and boost clocks up to 146 MHz (6.5%) faster. Techpowerup tested one of the faster OCed cards and found it was 4% more powerful overall than the stock card.
Yeah, I only realised that today when I saw videos from Steve Burke and Daniel Owen. I'm glad that it's better than I thought because, based on what Steve Walton said, it really was pointless.
 
Don't get me wrong I know it's a good deal. Everybody knows that value decreases as you go up the product stack. I didn't mean to say that it wasn't a good deal, it's just nothing new and exciting. It's kind of like how, sure the R7-5700X3D is a great deal, but there was no mystery to it at that point. I also was hampered by the fact that I had only seen Steve Walton's video and I didn't know until today that he did it wrong. Now that I've seen videos from Steve Burke and Daniel Owen, I realise that the card is in fact not pointless.

Yeah, I only realised that today when I saw videos from Steve Burke and Daniel Owen. I'm glad that it's better than I thought because, based on what Steve Walton said, it really was pointless.
Interestingly, Steve/Hardware Unboxed posted an updated review, taking into account the OCs. He does a good job of comparing stock 7800 XT to stock 7900 GRE, OC edition 7800 XT to OC edition 7900 GRE and stock 7800 XT to OCed 7900 GRE.

HW Unboxed 7900GRE review update

Basically, the OCed 7900 GREs are about 8% more powerful than the OCed 7800 XTs for that 10% extra cost. (And about 13% faster than the stock 7800 XT.) But factor in the fact that the OCed models of 7800 XT are generally more expensive than the MSRP for stock and the fact that at least a few models of OCed 7900 GREs are currently priced at the $550 MSRP, and the discrepancy between pricing and performance becomes basically zero, and the 7900 GRE becomes even more compelling, especially for people upgrading from previous GPU generations.

But the most interesting development in all of this is that all of the hubbub and dissatisfaction from much of the fanbase has forced AMD to announce that the strict limitations on overclocking that they obviously deliberately imposed on the 7900 GREs to avoid cannibalizing sales of the 7900 XT were actually a "bug" that will be fixed soon. So, we'll probably likely see those already decent OC performance numbers going up quite a bit with the next Adrenaline update.

So, in just a couple of days the enthusiast fanbase has managed to pressure both one of the largest YouTube hardware DUI enthusiast channels and the second largest computer parts manufacturer to change courses on some questionable decisions. Not a bad week for the community!
 
  • Like
Reactions: Avro Arrow
Interestingly, Steve/Hardware Unboxed posted an updated review, taking into account the OCs. He does a good job of comparing stock 7800 XT to stock 7900 GRE, OC edition 7800 XT to OC edition 7900 GRE and stock 7800 XT to OCed 7900 GRE.

HW Unboxed 7900GRE review update

Basically, the OCed 7900 GREs are about 8% more powerful than the OCed 7800 XTs for that 10% extra cost. (And about 13% faster than the stock 7800 XT.) But factor in the fact that the OCed models of 7800 XT are generally more expensive than the MSRP for stock and the fact that at least a few models of OCed 7900 GREs are currently priced at the $550 MSRP, and the discrepancy between pricing and performance becomes basically zero, and the 7900 GRE becomes even more compelling, especially for people upgrading from previous GPU generations.

But the most interesting development in all of this is that all of the hubbub and dissatisfaction from much of the fanbase has forced AMD to announce that the strict limitations on overclocking that they obviously deliberately imposed on the 7900 GREs to avoid cannibalizing sales of the 7900 XT were actually a "bug" that will be fixed soon. So, we'll probably likely see those already decent OC performance numbers going up quite a bit with the next Adrenaline update.

So, in just a couple of days the enthusiast fanbase has managed to pressure both one of the largest YouTube hardware DUI enthusiast channels and the second largest computer parts manufacturer to change courses on some questionable decisions. Not a bad week for the community!
Not bad at all. I understand why he used that method (it keeps things consistent) but he didn't take into account that nobody can currently buy a reference 7900 GRE because they were discontinued long ago. What puzzled me is that he used the Sapphire Nitro model when it was the Powercolor Hellhound that had the best clocks and thermals in his first video. I'm willing to bet that it's also a good deal less expensive than the Sapphire Nitro because the Nitro is Sapphire's marquee brand while the same cannot be said about Powercolor's Hellhound.

Powercolor tends to be all over the place with their branding. The only things that remain constant is that the Red Devil is their marquee brand and the Fighter is their base model. I'm honestly not sure if Red Dragon is higher than Hellhound or lower. Powercolor is a pretty inconsistent mess that way. 🤔
 
Powercolor tends to be all over the place with their branding. The only things that remain constant is that the Red Devil is their marquee brand and the Fighter is their base model. I'm honestly not sure if Red Dragon is higher than Hellhound or lower. Powercolor is a pretty inconsistent mess that way. 🤔
Red Devil
Red Dragon
Hellhound
Fighter

If it helps, one of these things is not like the others! The top three all have "evil / tough / monstrous" connotations. And then we have the lowly and generic Fighter. Also, I think Red Devil and Red Dragon are mostly reserved for higher tier products — like there's not an RX 7600 Red Dragon as an example. I don't think there are higher tier Fighter products either, for example.
 
  • Like
Reactions: Avro Arrow
Red Devil
Red Dragon
Hellhound
Fighter

If it helps, one of these things is not like the others! The top three all have "evil / tough / monstrous" connotations.
Thanks for that Jared! I was always a bit confused about the RD and HH.
And then we have the lowly and generic Fighter. Also, I think Red Devil and Red Dragon are mostly reserved for higher tier products — like there's not an RX 7600 Red Dragon as an example. I don't think there are higher tier Fighter products either, for example.
Oh yeah, the Fighter model is about as generic as it gets. Just a cheap back plastic shroud with fans. You know, now that I think about it, there actually is one model tier below Fighter in the Powercolor line, the ITX cards. I should've remembered it because I bought a Powercolor RX 6500 XT ITX for my mother's HTPC. It just looks like a single-fan Fighter model so Powercolor has the complete generic collection with that.

You can have one, two or three fans and they're all as plain-looking as possible:
7724-front.jpg

9441-front.jpg

11390-front.jpg

Having said that, it's not like we need more than what we see here because as cheap-looking as they are, they still work just fine. For those of us without glass side panels, having a "pretty" video card doesn't make a lick of difference, eh? 😉
 
  • Like
Reactions: JarredWaltonGPU
Status
Not open for further replies.