FRONTPAGE XFX MERC 310 RX 7900 XT Review: AMD’s Back in the Game

[Overclockers.com]

Thanks to XFX, we have the MERC 310 BLACK RX 7900 XT on the test bench today. This is our first GPU from AMD's RDNA3 line of cards. The RX 7900 XT is based on the NAVI 31 die. According to AMD, their new design delivers nearly 1.5x the performance per watt.

Click here to read more!

 

[JLK03F150]

I'm glad to see the AMD 7900 XT/XTX cards being competitive with the RTX 4080 for less money (well, time will tell on that point ).

Thanks for the review, Johan.

 

[EarthDog]

Yep... even though the price difference isn't much between the 4080 and 7900XTX ($200). I think NV has to come down just a bit on this card.... which I'd guess will make them think about the cost of the 4070/4070Ti coming in early Jan (depending on where it's performing).

I'm really surprised about the power use. It uses notably more power than the 4080 all around, but especially with multi-monitor and idling.

We see nice gains in ray tracing but still behind NV (if that feature is important to you).

 

[JLK03F150]

Even though RT has been implemented in MSFS2020, I'm not too concerned about it. It might make a difference in the Assetto Corsa series, but I'm just trying to not hit a wall or another car. I'm not looking at janky power lines overhead. lol

Yes, I'm hoping Nvidia lowers prices instead of AMD raising their prices to level the field.

 

[Culbrelai]

What a terrible showing tbh. RT performance of these new AMDs is between 3080 and 3090ti. Worse power consumption than 4080, powerdraw bug with multimonitors (100w at idle oof), awful coil whine on reference designs, and super duper overpriced AIBs, to the point where the AIBs are reaching into 4080 price territory.

Looks like Vega Part II.

 

[Hardin]

They should have replaced the Xs with $s. Having said that though thank you for this review. It's a lot more favorable to the card then the other reviews I've been seeing. It looks like you might have to be very selective about which AIB you go with this generation.

 

[Johan45]

I'm really surprised about the power use. It uses notably more power than the 4080 all around, but especially with multi-monitor and idling.

Me too; a couple of things bother me about it. The first thing is in any DX12 bench, while overclocked, the fans will spin up out of control (3300 RPM). This is not temperature related, and the fans will stay at 100% until I exit the benchmark. I have seen this before and always felt it was driver/CCC related, something to do with their interaction. Second, the amount of power when overclocked seems way too high. So far, GPUz doesn't show GPU power draw, so I did some benches afterward with the Radeon overlay that does show many of these stats. The Radeon Overlay was showing 400 W plus for the total board use of the GPU. Technically that shouldn't be possible. PCIe 8-pin is 150 W x2 is 300 W + 75 W for the slot so a total of 375 W. At least, that's the way I understand it.

 

[EarthDog]

Me too; a couple of things bother me about it. The first thing is in any DX12 bench, while overclocked, the fans will spin up out of control (3300 RPM). This is not temperature related, and the fans will stay at 100% until I exit the benchmark. I have seen this before and always felt it was driver/CCC related, something to do with their interaction. Second, the amount of power when overclocked seems way too high. So far, GPUz doesn't show GPU power draw, so I did some benches afterward with the Radeon overlay that does show many of these stats. The Radeon Overlay was showing 400 W plus for the total board use of the GPU. Technically that shouldn't be possible. PCIe 8-pin is 150 W x2 is 300 W + 75 W for the slot so a total of 375 W. At least, that's the way I understand it.

GPUz 2.51 was just released, maybe that has it?

Those numbers are the specs for the connectors. Most power supplies don't have anything stopping them from drawing at least a little bit more than that. Remember the 500W R9 295x2? That thing had the same config IIRC.

 

[Johan45]

GPUz 2.51 was just released, maybe that has it?

Those numbers are the specs for the connectors. Most power supplies don't have anything stopping them from drawing at least a little bit more than that. Remember the 500W R9 295x2? That thing had the same config IIRC.

Yes, but even according to the GPU specs it "shouldn't" be drawing that much power. With a +15% option, it still shouldn't be over 375W unless that's just a rough guide? Usually, we're fighting to get more power allotment.

That's the GPUz I used for the review

 

[Johan45]

any DX12 bench, while overclocked, the fans will spin up out of control (3300 RPM). This is not temperature related, and the fans will stay at 100% until I exit the benchmark. I have seen this before and always felt it was driver/CCC related, something to do with their interaction.

Ok, so I found this in the lit that XFX sent me:

Manually adjustments to the fan via the AMD Adrenalin Software will override the BIOS, a full driver wipe and DDU may be required to restore the factory BIOS fan profile any time a manual profile is used.

Since I do/will/did fiddle with the fan settings while overclocking, I followed this procedure, used DDU in safe mode, and reloaded the driver. Reran SOTR with the GPU overclocked, leaving the fans on auto, and had the exact same experience. The fans spun up and sat at 100% until I exited the game altogether. What's odd, though, is it's only while overclocked. If I reset the card to stock, it'll run fine, and the fans are barely audible.
I also tried switching from an Intel platform to a complete AMD platform, but there's no change.