3090 Low performance, not sure why, need help

[BringBackGPUMascots]

Hey y'all, ever since I got a 3090 it's been underperforming compared to all the benchmarks, and I've tried many things to fix it but I'm still getting 15%+ performance loss compared to others.
For instance, Baldur's Gate 3 runs at under 120FPS on Ultra, when techpowerup's new bench states 175.
I don't have the best CPU (Ryzen 5700X) but compared to my friend that has a 6900XT and a 5700X, he still gets ~30FPS more than me in some games.
I've got no clue what to do anymore, what am I missing?

This is all on 1080p, so I understand the CPU matters, I don't expect MAX performance.

Things I've tried:
BIOS updates, different GPU drivers, chipset drivers, Windows is fully up to date(WIN 10)
Undervolting the GPU, and leaving it stock, same FPS.
Upgraded CPU from 3600X to 5700X, tried PBO2(works fine, i get 4.6GHz all-core)
My GPU temps are always around 65-70°C when gaming, memory temps are ~75°C
Removing all background programs(MSI AB gave me a few FPS extra after shutting it down)

PC Specs are:
Ryzen 7 5700X @ Stock with PBO2 on Auto
Gigabyte AB350M GAMING 3
MSI RTX 3090 GAMING X TRIO
32GB GSKILL/CORSAIR DDR4 3000MHz CL16
Windows 10 Pro 22H2 build 19045.3208
GPU Drivers: 536.99 Game-Ready


What else can I try? I could settle for 110FPS in all games, but this thing was a bit too expensive to "settle".

Thanks!

 

[oxrufiioxo]

Hey y'all, ever since I got a 3090 it's been underperforming compared to all the benchmarks, and I've tried many things to fix it but I'm still getting 15%+ performance loss compared to others.
For instance, Baldur's Gate 3 runs at under 120FPS on Ultra, when techpowerup's new bench states 175.
I don't have the best CPU (Ryzen 5700X) but compared to my friend that has a 6900XT and a 5700X, he still gets ~30FPS more than me in some games.
I've got no clue what to do anymore, what am I missing?

This is all on 1080p, so I understand the CPU matters, I don't expect MAX performance.

Things I've tried:
BIOS updates, different GPU drivers, chipset drivers, Windows is fully up to date(WIN 10)
Undervolting the GPU, and leaving it stock, same FPS.
Upgraded CPU from 3600X to 5700X, tried PBO2(works fine, i get 4.6GHz all-core)
My GPU temps are always around 65-70°C when gaming, memory temps are ~75°C
Removing all background programs(MSI AB gave me a few FPS extra after shutting it down)

PC Specs are:
Ryzen 7 5700X @ Stock with PBO2 on Auto
Gigabyte AB350M GAMING 3
MSI RTX 3090 GAMING X TRIO
32GB GSKILL/CORSAIR DDR4 3000MHz CL16
Windows 10 Pro 22H2 build 19045.3208
GPU Drivers: 536.99 Game-Ready


What else can I try? I could settle for 110FPS in all games, but this thing was a bit too expensive to "settle".

Thanks!

Run the Timespy benchmark its free and will let you know if your general performance is in the normal range. It's not surprising that the 6900XT is running faster at 1080p in a lot of games amd has less cpu overhead. Your ram is pretty slow and could also be losing you 5-10% performance also mixing two different kits your timings are likely terrible.

The BG3 Benchmark on this site was done with a 13900K which is substantially better than your 5700X.

 

[BringBackGPUMascots]

Run the Timespy benchmark its free and will let you know if your general performance is in the normal range. It's not surprising that the 6900XT is running faster at 1080p in a lot of games amd has less cpu overhead. Your ram is pretty slow and could also be losing you 5-10% performance also mixing two different kits your timings are likely terrible.

The BG3 Benchmark on this site was done with a 13900K which is substantially better than your 5700X.

Is it the

3DMark Download v2.25.8043 + Time Spy?​

from guru3d? That's the only free one, the other one on Steam is 35$.

RAM timings seem to be as advertised on the ram sticks. I'll give it a go and report back what it tells me, thanks.

 

[oxrufiioxo]

Is it the

3DMark Download v2.25.8043 + Time Spy?​

from guru3d? That's the only free one, the other one on Steam is 35$.

RAM timings seem to be as advertised on the ram sticks. I'll give it a go and report back what it tells me, thanks.

Thats the one and dont mess with your ram timings unless you are familiar with tuning memory you could end up with frequent bsod or corrupt windows..... Maybe @tabascosauz could help with that but mixing kits will make that pretty difficult.

 

[tabascosauz]

Is it the

3DMark Download v2.25.8043 + Time Spy?​

from guru3d? That's the only free one, the other one on Steam is 35$.

RAM timings seem to be as advertised on the ram sticks. I'll give it a go and report back what it tells me, thanks.

Thats the one and dont mess with your ram timings unless you are familiar with tuning memory you could end up with frequent bsod or corrupt windows..... Maybe @tabascosauz could help with that but mixing kits will make that pretty difficult.

Yeah that's probably a no from me dawg......not that I don't want to help but im not exactly a mem OC whiz, and there are probably none out there who focus on optimizing absolute dumpster fire RAM kits like this

There probably isn't a single reason behind the low perf, more a combination of them:

• Even at release 6900XT performed the best in relation to 3090 at lower resolutions
• 6900XT probably finewined a bit in the past few years
• Radeon has less driver overhead, which potentially matters most at lower resolutions and on weaker CPUs (which the 5700X most definitely is, at this point in 2023 for 1080p)
• If 3200CL16 is what we usually term garbage, then 3000CL16 is beyond garbage
• Low Fabric clock of 1500MHz because 3000CL16
• Mixing kits, mystery timings, take a screenshot of https://zentimings.protonrom.com/
• Did you ever even memtest stable? Run 1usmusv3 config real quick https://www.overclock.net/threads/memory-testing-with-testmem5-tm5-with-custom-configs.1751608/
• Old-ass crappy board is a complete unknown as to whether latest BIOS/AGESA offers decent performance
• Old-ass crappy board is a complete unknown as to whether trash tier VRM even handles the 5700X at 76W sustained

The frametime graphs don't look out of place, really. There are some spikes here and there but the fact that you play at 1080p just sets you up for failure, even if I was more familiar with the performance of those specific games I probably wouldn't blame the game. 1080p is a rough time all around if you care about 1% lows + you have a 5700X + you have Geforce.

Run some Timespy.

 

[oxrufiioxo]

Yeah that's probably a no from me dawg......not that I don't want to help but im not exactly a mem OC whiz, and there are probably none out there who focus on optimizing absolute dumpster fire RAM kits like this

There probably isn't a single reason behind the low perf, more a combination of them:

• Even at release 6900XT performed the best in relation to 3090 at lower resolutions
• 6900XT probably finewined a bit in the past few years
• Radeon has less driver overhead, which potentially matters most at lower resolutions and on weaker CPUs (which the 5700X most definitely is, at this point in 2023 for 1080p)
• If 3200CL16 is what we usually term garbage, then 3000CL16 is beyond garbage
• Low Fabric clock of 1500MHz
• Mixing kits, mystery timings, take a screenshot of https://zentimings.protonrom.com/
• Did you ever even memtest stable? Run 1usmusv3 config real quick https://www.overclock.net/threads/memory-testing-with-testmem5-tm5-with-custom-configs.1751608/
• Old-ass crappy board is a complete unknown as to whether latest BIOS/AGESA offers decent performance
• Old-ass crappy board is a complete unknown as to whether trash tier VRM even handles the 5700X at 76W sustained

The frametime graphs don't look out of place, really. There are some spikes here and there but the fact that you play at 1080p just sets you up for failure, even if I was more familiar with the performance of those specific games I probably wouldn't blame the game.

Run some Timespy.

It was still very beneficial to tag you, You pretty much summed up everything I was thinking much better than I could when it pertains to why his performance is likely slightly lower than he expects.

I would also add it's pretty safe to ignore any YouTube/website benchmarks unless they have identical hardware which they will not.

A 5800X3D likely would have been the better bet with your setup although given it likely cost twice what you paid for the 5700X maybe not.....

 

[tabascosauz]

It was still very beneficial to tag you, You pretty much summed up everything I was thinking much better than I could when it pertains to why his performance is likely slightly lower than he expects.

I would also add it's pretty safe to ignore any YouTube/website benchmarks unless they have identical hardware which they will not.

A 5800X3D likely would have been the better bet with your setup although given it likely cost twice what you paid for the 5700X maybe not.....

Still, a 5800X3D potentially makes the most difference at 1080p. That, and X3D generally levels the playing field as to slow RAM and low Fabric speeds. Though, VRM becomes a major concern on the AB350M Gaming CF as the 5800X3D is nominally a 100-120W part, not 76W.

 

[oxrufiioxo]

Still, a 5800X3D potentially makes the most difference at 1080p. That, and X3D generally levels the playing field as to slow RAM and low Fabric speeds. Though, VRM becomes a major concern on the AB350M Gaming CF as the 5800X3D is nominally a 100-120W part, not 76W.

Yeah, but even capped at 75w it should destroy a 5700X in a lot of scenarios....

 

[BringBackGPUMascots]

Thats the one and dont mess with your ram timings unless you are familiar with tuning memory you could end up with frequent bsod or corrupt windows..... Maybe @tabascosauz could help with that but mixing kits will make that pretty difficult.

Okay these are the scores(in attachments), and from looking up others' CPU scores, they get around 12000+, when i get only 10000. I see their 5700X-es stay at 4.6GHz, while TimeSpy tells me my Max clock is 4.6, but my average is 3.8.
Only thing that would make sense here is their Average Temps are ~60°C, my average CPU temp is 71°C(often goes to 80-ish).

saving space by editing

I guess I'm just going by random online reviews, all of them say most of the time the 3090 is better, but yeah, probably like you said, AMD GPUs age like wine, hah.
Running that Testmem config now, i ran a different one by accident but it didn't show any errors. Zentimings are in attachments, I honestly don't really want to mess with RAM OC-ing, it's a bit out of my scope.

It was still very beneficial to tag you, You pretty much summed up everything I was thinking much better than I could when it pertains to why his performance is likely slightly lower than he expects.

I would also add it's pretty safe to ignore any YouTube/website benchmarks unless they have identical hardware which they will not.

A 5800X3D likely would have been the better bet with your setup although given it likely cost twice what you paid for the 5700X maybe not.....

I paid 130$(after selling my 3600X) for the 5700X new(in an actual legit store IRL, not aliexpress), so while I feel like I got a good deal, not sure if it was worth it...

5800X3D is out of my price range, cause it would be ~300$. And I'd need a new PSU for it, I'm on a Seasonic 650W 80+G.

 

[Dr. Dro]

Your system's too slow to keep up with TPU's 13900K test bench. The issues you have here are as follows:

- Slower CPU
- Very low-end motherboard limits the upper clock ceiling of your CPU, it won't boost as high
- Much slower RAM with much higher latency (which limits your CPU's full potential, most Ryzen test benches are using DDR4-3600 memory with 1800 FCLK, you're at 3000/1500)
- PCIe 3.0 platform to handle a terabyte-class bandwidth GPU (this normally doesn't count a lot - but it's a 3090 moving up to 1 TB of data per second bidirectionally, so expect a minor perf loss here, already adding to the numerous factors that make your machine slower than TPU's test bench) - 6900 XT exhibits a little less of a hit because it relies on its cache a lot more than raw data throughput as the 3090 does

I'd say if you bought a better motherboard that could flex your CPU's muscles a little bit more and was PCIe 4.0 capable (so a B550 or X570) and tune your RAM (try getting it to 3600, if not possible, try 3200 tightened), much of this discrepancy between your machine and your friends would disappear.

That said, I don't think it's worth investing a lot of money on this system, see if you can buy a B550 board for $150 tops and sell off your current board to make back some of the expense, but that'd be about all you can do.

 

[oxrufiioxo]

Okay these are the scores(in attachments), and from looking up others' CPU scores, they get around 12000+, when i get only 10000. I see their 5700X-es stay at 4.6GHz, while TimeSpy tells me my Max clock is 4.6, but my average is 3.8.
Only thing that would make sense here is their Average Temps are ~60°C, my average CPU temp is 71°C(often goes to 80-ish).


I guess I'm just going by random online reviews, all of them say most of the time the 3090 is better, but yeah, probably like you said, AMD GPUs age like wine, hah.
Running that Testmem config now, i ran a different one by accident but it didn't show any errors. Zentimings are in attachments, I honestly don't really want to mess with RAM OC-ing, it's a bit out of my scope.

I paid 130$(after selling my 3600X) for the 5700X new(in an actual legit store IRL, not aliexpress), so while I feel like I got a good deal, not sure if it was worth it...

5800X3D is out of my price range, cause it would be ~300$. And I'd need a new PSU for it, I'm on a Seasonic 650W 80+G.

Honestly you're not far off the average for your system. It's most likely performing as it should given your specs.


Better memory and tuned cpus on better motherboards is likely the majority of the performance difference you are seeing. Memory makes a massive difference on Zen 1-4. Honestly outside of a 5800X3D or buying substantially better low latency 3600 memory there is likely not much else you could do and even then your motherboard might hold you back a bit.

Your performance overall is still pretty good try running with DSR and enjoy better image quality likely at a similar framerate.

 

[Dr. Dro]

Your performance overall is still pretty good try running with DSR and enjoy better image quality likely at a similar framerate.

Yup, I'd say this is the best course of action if OP doesn't mind being a bit behind on the performance curve after all. The slow RAM + low-end motherboard are what's limiting them, but anything more than a modest motherboard upgrade and RAM tweaking is too much to spend on an AM4 system at this point if budget is of any concern.

 

[BringBackGPUMascots]

Here's TestMem, I have no clue what these numbers mean, but I was actively using the PC while it was running, so eh.

Your system's too slow to keep up with TPU's 13900K test bench. The issues you have here are as follows:

- Slower CPU
- Very low-end motherboard limits the upper clock ceiling of your CPU, it won't boost as high
- Much slower RAM with much higher latency (which limits your CPU's full potential, most Ryzen test benches are using DDR4-3600 memory with 1800 FCLK, you're at 3000/1500)
- PCIe 3.0 platform to handle a terabyte-class bandwidth GPU (this normally doesn't count a lot - but it's a 3090 moving up to 1 TB of data per second bidirectionally, so expect a minor perf loss here, already adding to the numerous factors that make your machine slower than TPU's test bench) - 6900 XT exhibits a little less of a hit because it relies on its cache a lot more than raw data throughput as the 3090 does

I'd say if you bought a better motherboard that could flex your CPU's muscles a little bit more and was PCIe 4.0 capable (so a B550 or X570) and tune your RAM (try getting it to 3600, if not possible, try 3200 tightened), much of this discrepancy between your machine and your friends would disappear.

That said, I don't think it's worth investing a lot of money on this system, see if you can buy a B550 board for $150 tops and sell off your current board to make back some of the expense, but that'd be about all you can do.

Yeah I just wanted to get the max i could for the next year or two, and I got the 3090 for REALLY cheap(650$, and prices here in Europe are insane, a new one is 1500$)
This is the board I had with a Ryzen 1600 (6?)years ago, it seemed to be still kicking fine, so I wanted to get a 120FPS@1440p machine out of it, and then probably upgrading to an 8800X3D when they come out.
I really like the 3090.

Honestly you're not far off the average for your system. It's most likely performing as it should given your specs.


Better memory and tuned cpus on better motherboards is likely the majority of the performance difference you are seeing. Memory makes a massive difference on Zen 1-4. Honestly outside of a 5800X3D or buying substantially better low latency 3600 memory there is likely not much else you could do and even then your motherboard might hold you back a bit.

Your performance overall is still pretty good try running with DSR and enjoy better image quality likely at a similar framerate.

Okay, do you think I could get a tiny bit more if i bought a better CPU cooler(I'm on a Wraith Spire now, from my 1600) I could OC the thing properly and get to that 120FPS@1440p number?
DSR is pretty great, though I haven't tested it much.

And of course, thank you all for helping!

 

[oxrufiioxo]

Here's TestMem, I have no clue what these numbers mean, but I was actively using the PC while it was running, so eh.

Yeah I just wanted to get the max i could for the next year or two, and I got the 3090 for REALLY cheap(650$, and prices here in Europe are insane, a new one is 1500$)
This is the board I had with a Ryzen 1600 (6?)years ago, it seemed to be still kicking fine, so I wanted to get a 120FPS@1440p machine out of it, and then probably upgrading to an 8800X3D when they come out.
I really like the 3090.

Okay, do you think I could get a tiny bit more if i bought a better CPU cooler(I'm on a Wraith Spire now, from my 1600) I could OC the thing properly and get to that 120FPS@1440p number?
DSR is pretty great, though I haven't tested it much.

It definitely couldn't hurt...

This is my secondary with similar specs to yours but with better memory/motherboard and a 5800X and it's honestly about the best you could expect from your system if you tweaked everything and bought better memory/motherboard and honestly it wouldn't be worth it.

I scored 18 695 in Time Spy

AMD Ryzen 7 5800X, NVIDIA GeForce RTX 3080 Ti x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

Your memory does seem like it isn't playing nice together try each kit separately and see if you get better performance with one.

 

[Dr. Dro]

Here's TestMem, I have no clue what these numbers mean, but I was actively using the PC while it was running, so eh.

Yeah I just wanted to get the max i could for the next year or two, and I got the 3090 for REALLY cheap(650$, and prices here in Europe are insane, a new one is 1500$)
This is the board I had with a Ryzen 1600 (6?)years ago, it seemed to be still kicking fine, so I wanted to get a 120FPS@1440p machine out of it, and then probably upgrading to an 8800X3D when they come out.
I really like the 3090.

Okay, do you think I could get a tiny bit more if i bought a better CPU cooler(I'm on a Wraith Spire now, from my 1600) I could OC the thing properly and get to that 120FPS@1440p number?
DSR is pretty great, though I haven't tested it much.

OOF, that testmem looks horribad. I'm surprised your machine isn't crashing apps left and right! I have a suspicion one of your sticks are bad, since you don't seem to be overclocking anything. Test your sticks one by one until you find the bad one and take it out... or take the opportunity to buy a 32 GB DDR4-3600 kit, they're super cheap these days, you may find one in the $60-80 USD range without much difficulty.

And yea, I agree, that was a great deal. The 3090 is still an awesome card, very powerful. I sold mine for about that much, too (a TUF OC). New owner seemed happy. Let's hope you can solve these issues and enjoy your system too!

 

[BringBackGPUMascots]

It definitely couldn't hurt...

This is my secondary with similar specs to yours but with better memory/motherboard and a 5800X and it's honestly about the best you could expect from your system if you tweaked everything and bought better memory/motherboard and honestly it wouldn't be worth it.

I scored 18 695 in Time Spy

AMD Ryzen 7 5800X, NVIDIA GeForce RTX 3080 Ti x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

Your memory does seem like it isn't playing nice together try each kit separately and see if you get better performance with one.

I can't, I need 32 gigs otherwise games often crash, and the GSKILL kit i bought had the exact same specs as my old Corsairs, and was very cheap, so welp.
That's what I'm thinking too, no point in buying better now, except the CPU cooler.
Getting even slightly close to your CPU score would be lovely though.

OOF, that testmem looks horribad. I'm surprised your machine isn't crashing apps left and right! I have a suspicion one of your sticks are bad, since you don't seem to be overclocking anything. Test your sticks one by one until you find the bad one and take it out... or take the opportunity to buy a 32 GB DDR4-3600 kit, they're super cheap these days, you may find one in the $60-80 USD range without much difficulty.

And yea, I agree, that was a great deal. The 3090 is still an awesome card, very powerful. I sold mine for about that much, too (a TUF OC). New owner seemed happy. Let's hope you can solve these issues and enjoy your system too!

I haven't had a single crash of anything since I got them, and I do a lot of multitasking, stuff like rendering, 3D, games, CAD... it DID crash one of my Firefox tabs while it was running the test though!
I'm amazed how well this MSI is cooled, memory never goes over 80°C if I'm doing AI stuff or rendering, and even at 75% fans and it being right next to me on top of my table without a case, I can't hear it.
Guess I'll save it, sell the rest when the 8xxx Ryzen series come out.

What if I did decide to try and OC the RAM? As a... learning experience, of sorts, hah. Is it hard?
Okay I just checked, my MB supports 3200MHz, the GSKILLs run at 3000 but the Corsairs run at 3200 natively(checked that again too, i was wrong before). Soooo could I try OCing the GSKILLs by those 200MHz easily?

 

[tabascosauz]

I can't, I need 32 gigs otherwise games often crash, and the GSKILL kit i bought had the exact same specs as my old Corsairs, and was very cheap, so welp.
That's what I'm thinking too, no point in buying better now, except the CPU cooler.
Getting even slightly close to your CPU score would be lovely though.

I haven't had a single crash of anything since I got them, and I do a lot of multitasking, stuff like rendering, 3D, games, CAD... it DID crash one of my Firefox tabs while it was running the test though!
I'm amazed how well this MSI is cooled, memory never goes over 80°C if I'm doing AI stuff or rendering, and even at 75% fans and it being right next to me on top of my table without a case, I can't hear it.
Guess I'll save it, sell the rest when the 8xxx Ryzen series come out.

What if I did decide to try and OC the RAM? As a... learning experience, of sorts, hah. Is it hard?
Okay I just checked, my MB supports 3200MHz, the GSKILLs run at 3000 but the Corsairs run at 3200 natively(checked that again too, i was wrong before). Soooo could I try OCing the GSKILLs by those 200MHz easily?

No, because that's not how it works. You don't even know that they have the same ICs. Each IC behaves and OCs differently. If they even have any OC headroom at all, which is not a guarantee in this segment of the RAM market. It's not "hard", just a metric ton of trial and error and learning.

Mixing kits is only a reasonable idea if you have the same half-decent IC in both kits, relatively similar binning (read: quality of IC), and are aware of how the IC scales and tweaks.

Go back to 1 kit, and/or find any cheap 2x16GB 3600CL16/17/18 kit.

At a glance without knowing what the kits really are, the zentimings looks fine for a XMP, instability probably coming from mixing kits and/or dual rank on an especially weak 4DIMM 4 layer board. Whichever 2 sticks are in slots A2 and B2 are probably fine, but A1 and B1 may be super weak and struggling.

 

[BringBackGPUMascots]

No, because that's not how it works. You don't even know that they have the same ICs. Each IC behaves and OCs differently. It's not "hard", just a metric ton of trial and error and learning.

Mixing kits is only a reasonable idea if you have the same half-decent IC in both kits, relatively similar binning (read: quality of IC), and are aware of how the IC scales and tweaks.

Go back to 1 kit, and/or find any cheap 2x16GB 3600CL16/17/18 kit.

Okay, good to know. My motherboard specsheet and memory support list only goes up to 3200, so if that's the case I guess I'll just stick to what I've got and go the better CPU cooler/OC route.

Thank you all for the help, I greatly appreciate it.

 

[tabascosauz]

Okay, good to know. My motherboard specsheet and memory support list only goes up to 3200, so if that's the case I guess I'll just stick to what I've got and go the better CPU cooler/OC route.

Thank you all for the help, I greatly appreciate it.

Pretty sure you don't need a better cooler for a 5700X, regardless of what you're using. We were speculating about VRM cooling and performance which is not really something you can just change.

Can you send a Zentimings for both kits? You can select the DIMM slot to view in the dropdown at the bottom.

A 2x16GB kit, heck even a 3200CL16 kit, shouldn't be all that expensive these days. Even if it's still crappy ICs, it won't be mixing kits, and won't have to fill what seem to be clearly worse A1 and B1 slots.

 

[BringBackGPUMascots]

Pretty sure you don't need a better cooler for a 5700X, regardless of what you're using. We were speculating about VRM cooling and performance which is not really something you can just change.

Can you send a Zentimings for both kits? You can select the DIMM slot to view in the dropdown at the bottom.

A 2x16GB kit, heck even a 3200CL16 kit, shouldn't be all that expensive these days. Even if it's still crappy ICs, it won't be mixing kits, and won't have to fill what seem to be clearly worse A1 and B1 slots.

I'm not sure really, Max I ever got it to was 84C, and AMD says the limit is 90.

Right, my bad, here are all four slots.

 

[Tatty_Two]

A 650w power supply, even a decent one seems a little on the light side for an RTX3090, our review of the Gaming X Trio shows a peak gaming power draw of 406w, these cards can spike all the way up to 550w before the cards power limit has a chance to kick in.

 

[BringBackGPUMascots]

A 650w power supply, even a decent one seems a little on the light side for an RTX3090, our review of the Gaming X Trio shows a peak gaming power draw of 406w, these cards can spike all the way up to 550w before the cards power limit kicks in.

That was one of my worries too, and why I went for the 5700X in the end. Reviews say it eats up 70-90W, plus the average 320W for the GPU with my undervolt, plus let's say 75W for the motherboard/SSDs, should be fine, even if it isn't in the 50-70% range.

 

[oxrufiioxo]

As long as you're not having random system shutdowns it should be fine.

 

[Tatty_Two]

That was one of my worries too, and why I went for the 5700X in the end. Reviews say it eats up 70-90W, plus the average 320W for the GPU with my undervolt, plus let's say 75W for the motherboard/SSDs, should be fine, even if it isn't in the 50-70% range.

You may be OK but even the most relaxed GPU manufacturers minimum spec them to 750W, some 850W for overclocked models, maybe run a demanding game with HWinfo64 in the background (sensors) and check for Performance limit power, reliability voltage and the PSU's 12V rail just to exclude the PSU.

 

[BringBackGPUMascots]

You may be OK but even the most relaxed GPU manufacturers minimum spec them to 750W, some 850W for overclocked models, maybe run a demanding game with HWinfo64 in the background (sensors) and check for Performance limit power, reliability voltage and the PSU's 12V rail just to exclude the PSU.

I oftentimes run the GPU at 100% with around 17GB of VRAM usage and follow it closely with GPU-Z, it never really goes over 340W max, PerfCapReasons seem fine, but it doesnt seem to go over 91% TDP, even though the power limit is at 100% in MSI AB.