Soldato
Yes thanks for the input, I have spares to think about and room in the case, if the plate is warm to touch (like 70c) then even laying a slab of heatsink on top would transfer off some heat allowing direct fans to move the heat... hmm
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
30xx Series Founders Edition
- Thread starter Dirk Diggler
- Start date
Just to update I e-mailed Nvidia previously asking if it affected warranty to change & add additional thermal pads to the backside of the 3080 FE and they said it was find to do so as long as I didn't damage card.
Fast forward to e-mail their support asking for pad type and thickness for the backside of the card and them refusing to give me the sizes saying Nvidia wont supply me with this information, when I directed to them to my previous e-mail to someone different and I was just ensuring the right thickness pads and of equal or higher quality were used they said ANY modifications to the card or even opening of the card would invalidate warranty. I then pointed them to the previous e-mail in which they said the information I was given was incorrect and to return my card if it was "faulty". I pointed out it was a known issue (links to various articles) and I was just trying to potentially stop a premature failure of the card without a lengthy RMA process.
Fast forward to e-mail their support asking for pad type and thickness for the backside of the card and them refusing to give me the sizes saying Nvidia wont supply me with this information, when I directed to them to my previous e-mail to someone different and I was just ensuring the right thickness pads and of equal or higher quality were used they said ANY modifications to the card or even opening of the card would invalidate warranty. I then pointed them to the previous e-mail in which they said the information I was given was incorrect and to return my card if it was "faulty". I pointed out it was a known issue (links to various articles) and I was just trying to potentially stop a premature failure of the card without a lengthy RMA process.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 29,867
As expected, they'd bounce you around if you did mod the card and needed to RMA. Not worth it imo.
As usual, 'at own risk'.
As usual, 'at own risk'.
Associate
Nvidia really need to put out a statement on this, I see no reason why removing the back plate or repasting the GPU core should invalidate warranty. But if they say its ok then of course they open the flood gates to the novice or the inept owner damaging the card doing so and send for RMA knowing they have damaged the card.
Nvidia probably take the stance they are happy to RMA untouched cards all day long for poor thermal's on pads or anything else but don't want to be checking cards for user damage from re paste or thermal pad replacement. Yet we know doing so will improve thermals and most likely extend the cards life.
In the end i guess it will be at the owners risk if they want to improve thermal's on there card. the 3090 can get toasty for sure.
Soldato
I agree. The whole point on right to repair movement was this activity is not really bodging the card its actually maintenance of something you bought. Yes I understand there are some wally's out there but as long as you disclose the 'competent' aspect then they should honour the maintenance aspect even if its only for the duration of the warranty period.
How does it stand if you buy the GPU from the US where you have the that right but have it shipped to the uk?
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 29,867
They'd probably refuse it then you'd have to take them to court in the USA.
Soldato
For a card this scarce I've decided its not worth the risk. If the card went faulty and Nvidia then return it saying warranty invalidated as its been tampered with I would be sick to my stomach. Only safest route is to use an external fan/heatsink but again after a lot of reading it can get mixed results and not always worthwhile.
Associate
I think it has to be personal choice and confidence in doing it properly. Really if your careful and precise and don't rush you should be able to re paste gpu and replace thermal pads easy enough. if the card develops a fault not related to what you have done it becomes an issue. Some cards will fail no matter what you do. some will last many years more because
you re pasted the GPU or replaced the thermal pads. Most will push there cards to the edge to see what they can do, which is fine now and then, but run them 24/7 on the edge and they most likely will fail, just like running your car at the edge of the rev limit in every gear will reduce the life expectancy.
In the end its the user's choice if they want to improve cooling there card. placing a fan on the back plate is a good idea, but only if the heat transfers to the back plate with good thermal pads.
Last edited:
Soldato
If you’re gaming and not mining, and you run with an undervolt it’s not really a big problem.
I’m still testing to find my best undervolt curve settings, but at around 0.80v with a 3090Fe my vram temps are:
- quake II RTX: 98c
- control: max settings, max ray tracing, no DLSS: 94c. (With DLSS, mid 80s).
- other non ray tracing games: around the mid 80’s c
I believe gddr6x vram is rated good up to 110c, and with the fans on auto they don’t go above around 55% even with quake II temps, so the card must think the vram temps are ok?
Quake II RTX is the ultimate punisher I’ve found, so it’s the only thing I’m really using to test stability now with an undervolt.
And if you don’t undervolt, why aren’t you doing it?! Better performance for significantly less power and heat is a no brainier.
I’m still testing to find my best undervolt curve settings, but at around 0.80v with a 3090Fe my vram temps are:
- quake II RTX: 98c
- control: max settings, max ray tracing, no DLSS: 94c. (With DLSS, mid 80s).
- other non ray tracing games: around the mid 80’s c
I believe gddr6x vram is rated good up to 110c, and with the fans on auto they don’t go above around 55% even with quake II temps, so the card must think the vram temps are ok?
Quake II RTX is the ultimate punisher I’ve found, so it’s the only thing I’m really using to test stability now with an undervolt.
And if you don’t undervolt, why aren’t you doing it?! Better performance for significantly less power and heat is a no brainier.
Soldato
I forgot to post. I did use the hwinfo app and unless its not finished yet it measures the junction temp sensor but shows the same value for min max and average. This cant be right as the min should be the temp from cold boot. Mine shows 104 for them all.
I understand they were working on another sensor which would report the 'hotspot' temperature. At least when this is working you can do some before mod, after mod logs to see how you have improved the memory temps. It sounds like the accuracy of the temps reported are not very precise so waiting till some software uses this properly.
Mine hits a peak of 104c, which like you say is below the max and you would trigger the throttle sensor that kicks in for protection when you do.
On previous GPUs both nvidia and AMD dont seem to worry too much about memory temps judging by the coolers used and previous reviews constantly show AIB's running them quite hot.
I understand they were working on another sensor which would report the 'hotspot' temperature. At least when this is working you can do some before mod, after mod logs to see how you have improved the memory temps. It sounds like the accuracy of the temps reported are not very precise so waiting till some software uses this properly.
Mine hits a peak of 104c, which like you say is below the max and you would trigger the throttle sensor that kicks in for protection when you do.
On previous GPUs both nvidia and AMD dont seem to worry too much about memory temps judging by the coolers used and previous reviews constantly show AIB's running them quite hot.
Soldato
Soldato
I took a screen:
Thats not the latest version try V6.43.4365 which works fine for me (newest version to date is V6.43.4380 beta at current date)
Soldato
Oh right, when I tried updating see I got this so thought I was on latest version:
Ace, this is on the fresh one:
Replaced my 3090 FE back pads... Thermalright odysy 1.5mm
The installed pads from Nvidia may as well be crusty old toothpaste!
When mining (hobby not serious, 3090 is subpar choice vs 3080/3070, etc for mining), it can show up VRAM temp issue easily and predictably.
I can now do
Now
PL 85
-300 core
+1225 Mem
Fan speed 44/44
T-junction on VRAM max 100c now (Nvidia max is 110c
)
~121MH/s
Previously
PL 79
-300 core
+900 Mem
Fans 77/77
Tjunction 108c (with new thermal pads it's was 96c)
~105MH/s
First time ever doing any mod of any sort on a GPU
The installed pads from Nvidia may as well be crusty old toothpaste!
When mining (hobby not serious, 3090 is subpar choice vs 3080/3070, etc for mining), it can show up VRAM temp issue easily and predictably.
I can now do
Now
PL 85
-300 core
+1225 Mem
Fan speed 44/44
T-junction on VRAM max 100c now (Nvidia max is 110c
)~121MH/s
Previously
PL 79
-300 core
+900 Mem
Fans 77/77
Tjunction 108c (with new thermal pads it's was 96c)
~105MH/s
First time ever doing any mod of any sort on a GPU
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 29,867
Hope you all get on well in the Summer
Soldato
I almost wish HWinfo64 had never updated to allow us to see these Vram temps, as I'm way happy with my core temps - never goes over 69c.
I would have been ignorantly happy with this card. I still am for gaming but the Vram temp is as niggle.
If it gets to be a problem I think I have room to move the 3090FE to the bottom PCIE slot to allow some more room between the CPU cooler, and I have some old copper ram heatsinks and a noctua 90mm fan lying around.
Are there any reports of just doing this? (not removing backplate and changing thermal pads)
How many degrees C are / would be shaved off with adding little ram heatsinks to the backplate and a fan? Is it worth the effort and the compromised looks of the card?
(Not keen on opening up the card yet as it's a lot of cash to risk voiding the warranty)
I would have been ignorantly happy with this card. I still am for gaming but the Vram temp is as niggle.
If it gets to be a problem I think I have room to move the 3090FE to the bottom PCIE slot to allow some more room between the CPU cooler, and I have some old copper ram heatsinks and a noctua 90mm fan lying around.
Are there any reports of just doing this? (not removing backplate and changing thermal pads)
How many degrees C are / would be shaved off with adding little ram heatsinks to the backplate and a fan? Is it worth the effort and the compromised looks of the card?
(Not keen on opening up the card yet as it's a lot of cash to risk voiding the warranty)
Last edited:
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 29,867
LOL just ignore it and enjoy the card! If anything happens within 3 years then that's on nVidia, you won't get stung.