MSI GeForce RTX 5070 Gaming Trio OC

[W1zzard]

The MSI GeForce RTX 5070 Gaming Trio OC impresses with whisper-quiet noise levels, even with heavy gaming, and excellent temperatures. You also get a small factory overclock out of the box, which helps boost performance bit.

Show full review

 

[Denver]

Strange to remove the 9070—the supposed competitor to the 5070—from the charts.

 

[evernessince]

Strange to remove the 9070—the supposed competitor to the 5070—from the charts.

Yeah that's the first thing I looked for as it's the obvious comparison. Lot of people are going to be comparing the performance and considering the 12GB vs 16B VRAM difference.

 

[tommo1982]

Strange to remove the 9070—the supposed competitor to the 5070—from the charts.

I had similar thoughts.
I read a review on another website and RX9070 had similar performance to RTX5070, sometimes a little better. They had Palit RX5070 for review and for the same price one can buy XFX RX9070 XT, at least where I live.

 

[3DVCash]

Yikes, forget the 4090, the 5070 doesn't even match Ampere's flagship 5 years later. Embarrassing.

 

[wheresmycar]

thanks for the review!!

2% increase in perf for a mid-ranged card for a $100 more is kinda crazy. Then again, in the GPU world "Crazy" is just another way of saying "Monday"

Strange to remove the 9070—the supposed competitor to the 5070—from the charts.

It seems W1zzard forgot to add it to all subsequent charts since the 9070 review. It's missing in both 9700 XT reviews and yesterday's 5090 TUF review.

 

[docnorth]

Unfortunately even this price, 100$ higher than FE would be considered a ‘bargain’ in Europe.

@W1zzard a little nitpicking, video playback consumes less than idling?

 

[jabbadap]

Yeah that's the first thing I looked for as it's the obvious comparison. Lot of people are going to be comparing the performance and considering the 12GB vs 16B VRAM difference.

Strange to remove the 9070—the supposed competitor to the 5070—from the charts.

Don't know if w1zzard have reference clock rx 9070, he usually does not use oc models on the charts.

 

[Scircura]

a little nitpicking, video playback consumes less than idling?

The power consumption time-graph at the top of the page shows the card entering a temporary higher power state during both the idle and multi-monitor tests. So the average consumption at idle really was measured higher than video playback.

Seems like a problem with the card's firmware. I would be curious if this is something that the card does only once after booting, or if it repeatedly cycles.

Edit: I see an idle power blip in other reviews: 5080 FE review has it, 5080 Suprim too.

In the the January 5080 reviews idle power was around 18-22 W for both FE and all partner models, but in the March 5070 reviews, 5080 is claimed to idle at 14 W. I'm curious why. 5070 Ti is still listed at 21 W idle.

 

[W1zzard]

The power consumption time-graph at the top of the page shows the card entering a temporary higher power state during both the idle and multi-monitor tests. So the average consumption at idle really was measured higher than video playback.

Seems like a problem with the card's firmware. I would be curious if this is something that the card does only once after booting, or if it repeatedly cycles.

Edit: I see an idle power blip in other reviews: 5080 FE review has it, 5080 Suprim too.

In the the January 5080 reviews idle power was around 18-22 W for both FE and all partner models, but in the March 5070 reviews, 5080 is claimed to idle at 14 W. I'm curious why. 5070 Ti is still listed at 21 W idle.

For some reason the card jumps out of idle during the test, I'll rerun to confirm

http://imgur.com/RlP2DQ9

Don't know if w1zzard have reference clock rx 9070, he usually does not use oc models on the charts.

I don't, that's the problem. I guess I could just add the slowest 9070 for now

 

[Sake]

Don't know if w1zzard have reference clock rx 9070, he usually does not use oc models on the charts.

MSI GeForce RTX 5070 Gaming Trio OC, the tested model is also overclocked.

 

[gridracedriver]

@W1zzard Why do I never see the RX 9070 in the tables?

 

[W1zzard]

@W1zzard Why do I never see the RX 9070 in the tables?

As mentioned two posts above yours, I don't have a reference clocked RX 9070, only OC models. But I think adding the numbers of the lowest model in this article should be ok, at least until I have better data

 

[gridracedriver]

As mentioned two posts above yours, I don't have a reference clocked RX 9070, only OC models. But I think adding the numbers of the lowest model in this article should be ok, at least until I have better data

Sorry, i posted without reading all the comments first, thanks.

 

[leonavis]

the 9070 is missing in the efficiency charts

edit: oh it's already been noted and explained. I noticed it because it's supposed to be very efficient, so it seemed strange that it's not there

 

[Dahita]

Between the ridiculous amount of references of RTX 5070 (19!!!!!) while none of them is available, the cheaper boards and quality, the disengagement with AMD and the recent artificial raise in pricing...

I don't know what MSI is playing at, but the problem sure ain't scalpers.

(Left: Asus Right: MSI, both 5070 Ti)

 

[N/A]

Between the ridiculous amount of references of RTX 5070 (19!!!!!) while none of them is available, the cheaper boards and quality, the disengagement with AMD and the recent artificial raise in pricing...

19 cards, of which there are OC /non OC variants/ in half, one with 3X or 2X fan options, one in white, one in premium, gaming and one in budget.
And you can see why the numbers can escalate quickly.
And of course, the PCB size has to accommodate the cooler and the higher end power supply.
Why don't you show the 5070 reviewed here, those are big PCBs. the small one is probably reference and 2X fan in mind.

 

[rusty caterpillar]

Between the ridiculous amount of references of RTX 5070 (19!!!!!) while none of them is available, the cheaper boards and quality, the disengagement with AMD and the recent artificial raise in pricing...

I don't know what MSI is playing at, but the problem sure ain't scalpers.

(Left: Asus Right: MSI, both 5070 Ti)

View attachment 389940

What can I see is Heat by proximity on MSI more than on Asus. Both producers didn't place a VRAM block close to the left VRM, which is good.

The power consumption time-graph at the top of the page shows the card entering a temporary higher power state during both the idle and multi-monitor tests. So the average consumption at idle really was measured higher than video playback.

Seems like a problem with the card's firmware. I would be curious if this is something that the card does only once after booting, or if it repeatedly cycles.

Edit: I see an idle power blip in other reviews: 5080 FE review has it, 5080 Suprim too.

In the the January 5080 reviews idle power was around 18-22 W for both FE and all partner models, but in the March 5070 reviews, 5080 is claimed to idle at 14 W. I'm curious why. 5070 Ti is still listed at 21 W idle.

In idle, the card is clearly searching for the missing ROPS... checking with Nvidia servers periodically if are some ROP available in stock.

 

[Dahita]

What can I see is Heat by proximity on MSI more than on Asus. Both producers didn't place a VRAM block close to the left VRM, which is good.

Also the MSI black MOSFETs are all misaligned. They are cheaper. Less power phases (11 VS 19!). The heatsink is also worse. You can see the full review here:

 

[redeye]

OMG, had no idea that in cp2077, 4k that the 9070xt is slightly better than the 7900xtx. wow just wow...
but, a 9070xt does not have waterblocks for it… right? right? lol.

 

[N/A]

Also the MSI black MOSFETs are all misaligned. They are cheaper. Less power phases (11 VS 19!). The heatsink is also worse. You can see the full review here:

Is it just me or he points at the inductors and says mosfets multiple times, the black or gray rectangles. Personally I don't care about how many phases if they are using doublers and how perfectly they align, as it's not a beauty contest and it doesn't matter much if it works as intended but it's true that Ventus 3x is all messed up. 41 dBA is outrageous.

 

[veaseomat]

cool write up, thanks. got 3 for AI inferencing. 36gb allows me to use models finetuned to fit on 32gb and still have headroom to game on. i prefer 250w cards. 250w at 3090 performance is nuts imo.

 

[blinnbanir]

As mentioned two posts above yours, I don't have a reference clocked RX 9070, only OC models. But I think adding the numbers of the lowest model in this article should be ok, at least until I have better data

Is this not OC as well? I could understand that if this was a reference 5070 but it is an OC model? Maybe I am missing something but are there any reference 9070s?

 

[leonavis]

are there any reference 9070s

Sapphire PULSE, Powercolor Reaper, Acer Nitro and AsRock Challenger

 

[Scircura]

cool write up, thanks. got 3 for AI inferencing. 36gb allows me to use models finetuned to fit on 32gb and still have headroom to game on. i prefer 250w cards. 250w at 3090 performance is nuts imo.

This is an interesting setup and I'd love it if you made a new forum thread with your experience gaming+inferencing with this arrangement. Does it actually work? Do games run correctly? How's the LLM + game performance (quick and informal benchmarks)?