Thermal testing two different size Gigabyte 5070 Ti cards - huge differences

[BoggledBeagle]

I just posted this video:

PCBs in the cards seem to be the same, tha smaller just less occupated thanks to missing RGB.

With automatic fan control the SFF model runs at 2000 RPM and 71°C with 300W Furmark load, has some unpleasant fan ringing too.

 

[ir_cow]

Different models. Makes sense to me.

 

[BoggledBeagle]

The SFF card runs normal stress tests (Speedway, Nomad) 99%+ at 66°C and fans at 1700 RPM.

 

[tugrul_LordOfDrinks]

When I set maximum temperature of 4070 ventus 2x to 55Celcius, the 5070 solid oc is 2.3x faster at same temperature.

 

[webmanau]

The SFF card runs normal stress tests (Speedway, Nomad) 99%+ at 66°C and fans at 1700 RPM.

Same here. Sure, the gaming OC runs cooler due to its much bigger heatsinks but it is also costs a good bit more. I wear headphones anyway so the extra noise does not bother me. The Windforce can be flashed with the Gaming BIOS and still runs at 66°C with an adjusted fan curve.

 

[GrimReaper85]

But why does my Gaming OC run so hot? Even a little hotter than Windforce from OP.

250W, 1000 rpm - 77°C, VRAM 86
250W, 1200 rpm - 73°C, VRAM 82
250W, 1500 rpm - 67°C, VRAM 78
250W, 2000 rpm - 62°C, VRAM 72
270W, 1200 rpm - 78°C, VRAM 86
270W, 1500 rpm - 71°C, VRAM 80
270W, 2000 rpm - 66°C, VRAM 74
300W, 1500 rpm - 77°C, VRAM 84
300W, 2000 rpm - 71°C, VRAM 78
300W, 2300 rpm - 70°C, VRAM 76
340W, 1500 rpm - 85°C, VRAM 88
340W, 2000 rpm - 77°C, VRAM 82

I mean, it's not the worst card I had, I put it in temperature tier 3 out of 4, same tier as 3080 12 GB MSI Suprim X, 6900 XT PowerColor RedDevil Ultimate and 3060 Ti Gigabyte Aorus Elite. And these of course are good cards, some of the best from beginning of 2022. Yet OP's Gaming OC proves it should be even much better.

My Gigabyte cards are all over the place, in all 4 different temperature tiers. So quite a lottery.

Anyway, only MSI Vanguard and Suprim and ASUS Astral from now on, I've had it with cheap cards.

 

[BoggledBeagle]

How old is our card? I examined one Gaming OC card from a third week (serial 2503...) and it had its PCB significantly bent. I have not powered that card up, so I do not know how hot it could run.

Perhaps your card has some mechanical defect that caused the cooler not to sit well on the chip and memory? I also noticed that the cards have quite large gaps between RAM packages and the cooler, which may not be filled with the thermal gel.

Gigabyte graphic card - TIM GEL application failure?

After inspecting a Gigabyte RTX 5070 Ti Gaming OC card with unusually strong smell, about which I reported here: https://www.techpowerup.com/forums/threads/smelly-brand-new-graphic-cards-remanufactured.334706/ I discovered, that my particular sample (SN 250641092344) seems to have pretty large...

www.techpowerup.com

EDIT - I just remembered that the Gaming OC model has a vapour chamber, I think your VC may be defect. I think you should RMA the card.

 

[Chrispy_]

But why does my Gaming OC run so hot? Even a little hotter than Windforce from OP.

The ambient temperature where you live and your case cooling have even more of an impact on noise/heat/temperatures than the cooler itself.
Then, you've also got vBIOS differences and silicon lottery.

Unless you're in an air-conditioned room with plenty of case cooling that ensures both good intake of cold air to the GPU fans and also good extraction of hot air without recirculating any of the hot exhaust into the intakes of the GPU again, your temperatures are your temperatures. The exact same GPU will behave differently in a different case in a different room with different airflow and temperatures.

I'm running a 5070Ti Ventus 3X which is supposedly one of the noisiest and hottest 5070Ti models reviewed, but I'm running it at 250W in an HTPC case with dual-120mm intakes just a couple of inches away and pointed directly at the GPU fans and dual 120mm exhausts almost directly opposite, so it's getting very very fresh air and there's a strong directional airflow preventing any stale exhaust air from being recirculated by the GPU cooler. It's practically silent and runs very very cool indeed.

 

[BoggledBeagle]

The ambient temperature where you live and your case cooling have even more of an impact on noise/heat/temperatures than the cooler itself.

I was thinking to ask the poster about the ambient temperature, but then I presumed that he would have mentioned in his post, if he was in a 30°C room.

 

[GrimReaper85]

The ambient temperature where you live and your case cooling have even more of an impact on noise/heat/temperatures than the cooler itself.
Then, you've also got vBIOS differences and silicon lottery.

Unless you're in an air-conditioned room with plenty of case cooling that ensures both good intake of cold air to the GPU fans and also good extraction of hot air without recirculating any of the hot exhaust into the intakes of the GPU again, your temperatures are your temperatures. The exact same GPU will behave differently in a different case in a different room with different airflow and temperatures.

I'm running a 5070Ti Ventus 3X which is supposedly one of the noisiest and hottest 5070Ti models reviewed, but I'm running it at 250W in an HTPC case with dual-120mm intakes just a couple of inches away and pointed directly at the GPU fans and dual 120mm exhausts almost directly opposite, so it's getting very very fresh air and there's a strong directional airflow preventing any stale exhaust air from being recirculated by the GPU cooler. It's practically silent and runs very very cool indeed.

Sure, but I was in the same room with all my cards. I even did an Excel table of all my cards temperatures (there were 12) since end of 2020 and I can clearly see this one is not the best. Even compared to RX 9070 with only 1100 g and 50 mm cooler it is at least 5°C worse.
And by my experience, GPU cooling itself has the most impact on its temperature.

My room temperature was 24-25°C during this testing which was 20 minutes on each setting. Then once it was 25.7°C I opened the window and it was again 25.2°C.

Case fans are at 600 rpm, but like I said, if GPU were good, it would have 10°C less regardless if my room temperature is 25°C and even if I disabled all case fans.

I remember I got the same answer when Gigabyte 2060 SUPER had 82°C. ASUS Dual 2060 SUPER then had 70°C.

It's exactly 14 days since I ordered it, so best to start writing return email.

 

[tugrul_LordOfDrinks]

Which is better:

Zotac non-oc model on top of a glacier in Norway during winter

Asus tuf model in Lut Desert in Iran during summer

 

[GrimReaper85]

At least now I have future reference for what are too high temperatures.

BoggledBeagle's temperatures for Gigabyte Gaming OC RTX 5070 Ti:
250W, 1050 rpm - 60°C (17°C less than mine, doesn't really matter at this point if room temperature is 22°C or 25°C and I doubt anyone with AC has it set to 20°C. 15°C or 30°C, that we can agree on that it would make a huge difference, but still probably only 10°C)
300W, 1715 rpm - 57°C (again around 17°C less than mine, I only tested at 1500 and 2000, but 1700 rpm should be somewhere in between)

Return email sent.
Last Gigabyte card for a while.

Now good question would be if card without vapor chamber is a more safe purchase. But then again, those have too high temperature as is. But comparing mine to Windforce, still a little less.
To be exact, 4°C less at 250W and 1500 rpm and 2°C less at 300W and 1700 rpm.

 

[insanee420]

Should have my 5070 Ti Windforce here by friday it seems, might update here with temps later so we can have more results

Sure, but I was in the same room with all my cards. I even did an Excel table of all my cards temperatures (there were 12) since end of 2020 and I can clearly see this one is not the best. Even compared to RX 9070 with only 1100 g and 50 mm cooler it is at least 5°C worse.
And by my experience, GPU cooling itself has the most impact on its temperature.

My room temperature was 24-25°C during this testing which was 20 minutes on each setting. Then once it was 25.7°C I opened the window and it was again 25.2°C.

Case fans are at 600 rpm, but like I said, if GPU were good, it would have 10°C less regardless if my room temperature is 25°C and even if I disabled all case fans.

I remember I got the same answer when Gigabyte 2060 SUPER had 82°C. ASUS Dual 2060 SUPER then had 70°C.

It's exactly 14 days since I ordered it, so best to start writing return email.

Fan rpm itself is only part of the equation as fan efficiency wildly varies, fan A at 600rpm might perform better than fan B at 1000-1200rpm for example, and fan B at 600rpm might do close to nothing, especially since we don't know which case you're even using, some cases have terrible porosity (closed front/top etc) and in some cases even if you blast a flagship fan at 3k rpm it won't matter because air still won't get in (or out), so if the fan itself is not the best and you have additional obstruction from the case then maybe 600rpm isn't doing much, and if the intake path is mostly blocked then maybe even 100% rpm won't do much

Just commenting since I've only seem fan rpm, ambient temp and GPU models listed but the case is just as important, I've seen quite a few people who got >10c temp drops simply by replacing their hotbox cases, might not be your case (ha) but I thought it's worth adding to the discussion here, something I've seen happen as well is people running big GPUs on small cases and the fans were almost pressed against a solid PSU top shroud, same thing for vertical-mount where they were sometimes way too close to the glass and the GPU probably struggles to even pull air in those situations

 

[GrimReaper85]

be quiet 802. I have 11 SSDs so changing the case in next 10-15 years is out of the question anyway.
As for the case fan speed, 600 rpm is the best I can do without them bothering me.

RTX 4090 Gigabyte Gaming OC in same case had 10°C less. At 250W and 1500 rpm it was 57°C.
Sure cooler was a little heavier, 2004 g instead of 1768 g. But that can't bring temperature down by 10°C.
That's why I was saying GPU itself is the most important for GPU temperature. As I don't believe I could lower the temperature of my 5070 Ti by moving the PC in 15°C room and best possisble case in the world with case fans at 2000 rpm. Maybe by 8°C.

 

[insanee420]

be quiet 802

Then it definitely seems a bit high if you're using the mesh front, if running with the solid panel I'd say that's expected, might improve a bit with slightly higher case fan rpm if using the mesh front, while if using solid front it might not make much of a difference running at 0rpm or 3k rpm as most of the air is being obstructed anyway, and that'd leave your case with almost no fresh air intake

RTX 4090 Gigabyte Gaming OC in same case had 10°C less. At 250W and 1500 rpm it was 57°C.
That's why I was saying GPU itself is the most important for GPU temperature. As I don't believe I could lower the temperature of my 5070 Ti by moving the PC in 15°C room and best possisble case in the world with case fans at 2000 rpm. Maybe by 8°C.

Sure but that's a different GPU even if it's still named Gaming OC, much bigger die to dissipate heat and likely a better cooler overall, no doubt a >450w TDP GPU will run much cooler at 250w than a ~300w GPU at 250w, though I'd say close to 60c is pretty high considering it's running only at 250w, which makes me inclined to think you're running the solid front panel as there's this noticeable trend of high-ish temps, which would make sense if the front intakes are being obstructed, though if you're running just the mesh front and front-to-back airflow then I don't know what's going on

 

[Chrispy_]

I was thinking to ask the poster about the ambient temperature, but then I presumed that he would have mentioned in his post, if he was in a 30°C room.

I don't assume, because there are plenty of people posting here from the Asian or South American tropics who endure a 35C room all day every day.
In theory, you'd assume people who can afford gaming PCs can also afford air-conditioning but I know for a fact that it's not always how things pan out

 

[BoggledBeagle]

The room temperature for my measurements was about 20 °C. I also measured with the case opened, so the cards had access to air of this temperature.

 

[GrimReaper85]

How old is our card? I examined one Gaming OC card from a third week (serial 2503...) and it had its PCB significantly bent. I have not powered that card up, so I do not know how hot it could run.

Perhaps your card has some mechanical defect that caused the cooler not to sit well on the chip and memory? I also noticed that the cards have quite large gaps between RAM packages and the cooler, which may not be filled with the thermal gel.

Gigabyte graphic card - TIM GEL application failure?

After inspecting a Gigabyte RTX 5070 Ti Gaming OC card with unusually strong smell, about which I reported here: https://www.techpowerup.com/forums/threads/smelly-brand-new-graphic-cards-remanufactured.334706/ I discovered, that my particular sample (SN 250641092344) seems to have pretty large...

www.techpowerup.com

EDIT - I just remembered that the Gaming OC model has a vapour chamber, I think your VC may be defect. I think you should RMA the card.

2508

Then it definitely seems a bit high if you're using the mesh front, if running with the solid panel I'd say that's expected, might improve a bit with slightly higher case fan rpm if using the mesh front, while if using solid front it might not make much of a difference running at 0rpm or 3k rpm as most of the air is being obstructed anyway, and that'd leave your case with almost no fresh air intake

Solid front panel.
No doubt BoggledBeagle's conditions are better, but 17°C better? I doubt it.

If 4090 can't be compared for whatever reason (cooler is actually 74 mm compared to 70 mm, not much difference), then how did PowerColor Hellhound 7800 XT had 62°C at 265W and 1165 rpm? That's 14°C less than 5070 Ti.
I checked that linked thread and it's most likely because Gigabyte still doesn't know how to make GPU's after more than 20 years. Like half of their GPUs are bad.

My 2 examples at 270W and 2000 rpm (returned both):
Gigabyte Gaming OC RTX 3070 - 75°C
Gigabyte Aorus Master RTX 3070 - 86°C

PowerColor Red Devil 6900 XT for comparison had 66°C. And as I said, Hellhound 7800 XT had even less at only 1165 rpm which is a lot less than 2000 rpm.

 

[webmanau]

FWIW, got a Windforce OC here, flashed to Gaming OC BIOS. @ 350W, default fan curve it runs the fans just under 2200RPM and 72°C GPU temp. If I run the fans faster it sits at about 66°C. @ 250W it runs 60°C and 1880 rpm fans.

 

[GrimReaper85]

Same as BoggledBeagle's then, a couple degrees cooler than my Gaming OC.
So mine came from the factory with failed vapor chamber or some other shoddy assembly mistake. And I remember from PowerColor factory planet video that they test every card for 1 hour in 3DMark, so I don't know how that can happen unless Gigabyte decided to cut costs (and they always do that, that's why they are the cheapest manufacturer for 20 years and counting) and test only every 10th card.

 

[insanee420]

Solid front panel.
No doubt BoggledBeagle's conditions are better, but 17°C better? I doubt it.

I've seen differences in the 10-20c range solely due to case so it's still possible, he mentioned 20c ambient so let's say 4-5c from ambient and 10c from case and that's pretty much 17c with some margin of error

If 4090 can't be compared for whatever reason (cooler is actually 74 mm compared to 70 mm, not much difference), then how did PowerColor Hellhound 7800 XT had 62°C at 265W and 1165 rpm? That's 14°C less than 5070 Ti.

GPU thickness doesn't say much, and even with the same cooler the die size is completely different, thermal density is gonna be way lower on a bigger die, can't really do 1:1 comparisons from a 4090 to a 5070 Ti to a 7800 XT like that as they're different in every possible way and it's not as simple as it looks. AMD GPUs as per TPU reviews tend to report ridiculously low GPU temps (but high hotspot IIRC), I'm talking sub 50's temps so from 45c to 62c or from 60c to 77c would be roughly the same 17c or so difference you're seeing all the time

I'd definitely at least test just the mesh front cause yeah, air can't get through a solid wall no matter if it's at 600 or 100k rpm, it's solid and air simply won't get in, therefore there's pretty much no cold air available for the GPU so it'll either keep heating up until it throttles or simply reach an equilibrium at a temp way higher than it could be if it had any access to fresh air, there's a reason most people moved away from those hotbox designs to mesh cases, if you care about temperatures then you definitely don't want to suffocate your hardware, in your case it seems pretty simple as you can just alternate between solid panel vs mesh panel, so simply replace the front panel (or just remove the solid one and leave it completely open) and see what results you get

 

[GrimReaper85]

With front panel removed and case fans (2x 140 mm intake, 1x 140 mm exhaust) set to Full Speed (1130 rpm) and room temperature 23.0-23.5°C with opened window (not much more than if I were using AC, maybe 1°C) I get:
300W, 1500 rpm - 67°C, VRAM 74°C

So both exactly 10°C less than before.
OK, so no RMA for me then.

Card also works fine with solid panel and case fans at 600 rpm. With a more sane power setting.
UV 900 mV, +400 +2000, Silent BIOS - 75°C, VRAM 84°C, 1475rpm, max 287.5W, avg. 273W
This is worst case anyway, I am using Quake II RTX instead of MSI Kombustor. I retested with Kombustor and got 68°C, VRAM 68°C. So not so hard on VRAM and I will continue using Quake II RTX.

Oh, as for the other cards, most of them were still in Corsair 600T with mesh front panel and 200 mm front intake fan.
Only Palit GameRock 4080 end of 2022 and Gigabyte Gaming OC 4090 in beginning of 2023 were in this be quiet 802 case.
So for 4080 it was December instead of April and I only have one Port Royal benchmark numbers and not 20 min Quake II RTX like now. It was 8°C less than 5070 Ti, so with same conditions probably pretty much similar.

Sorry to doubt everyone. Yes, my methodology needs some work. It's probably better for consistency to test GPU temperatures like I did now with open front panel than with closed.

 

[AleXXX666]

I just posted this video:

PCBs in the cards seem to be the same, tha smaller just less occupated thanks to missing RGB.

With automatic fan control the SFF model runs at 2000 RPM and 71°C with 300W Furmark load, has some unpleasant fan ringing too.

not surprising me - sh*tbyte are using cheap crappy crackling fans already looong time. I purchased BRAND NEW 2070 SUPER 3x windforce - fans cracling right from purchasing. Sent it RMA - they told "all good" lmfao. Sold this BS out. Have seen a lot of 16/20 series with crappy fans with very small usage time - not surprising me, I could tell by fans "design" that they are STILL same, only change with OVERPRICED "aorus" series.

 

[_roman_]

GPU thickness doesn't say much

Same as loudness of the fan or build quality.

gigglebyte - enough of the topic

two different cards.

Recent topic also may point out to buy hardware quite fast after a review. There could be component swaps which reduce paramaters of the hardware

 

[GrimReaper85]

I measured with same settings as BoggledBeagle. My conditions are a little worse, 24°C instead of 20°C room temperature and he had opened side panel.

250W, 1050 rpm - 65.2°C, so 5°C more than BoggledBeagle's, seems OK. Plus he said gaming, mine is stress test with constant 250W.
300W, 1140 rpm - 71.8°C, 8.2°C more than BoggledBeagle's
300W, 1715 rpm - 64.8°C, 7.8°C more than BoggledBeagle's

Even with just cleaned dust on front panel, I now get with solid front panel and 600 rpm case fans:
300W, 1500 rpm - 75°C, VRAM 80°C, before it was 77°C, VRAM 84°C