HD7950\7970 owners, whats your VRM temps under max load with Heaven 3.0?

[Ket]

Title says it all really. Googling isn't really giving me a proper answer... seems like a lot of people don't measure VRM temps, which is rather worrying in itself. The main reason I'm asking is because the VRM cooling... well the cooling as a whole really, on the HD7950 Vapor-X I bought was nothing short of abysmal, even after several attempts at re-seating the VRM & Vapor-X cooling my temps were still absolute crap.. 90c GPU temps and 106-120c VRM temps under load - this was all at stock frequencies too. After much re-designing of the VRM cooling and replacing the Vapor-X GPU cooler entirely with a Gelid Icy Vision A I think I have finally got all temps under control but would like to verify.

Heaven 3.0 Settings:

Extreme Tessellation
Shaders @ High
16x AF
8x AA
1680*1050 Resolution

CCC Settings:

Bone stock, except I changed Catalyst AI texture filtering quality to "Very High".

My temps after letting Heaven 3.0 run for about 20 mins:



Does that seem about right to everyone? Difficult to know what to set the bar at when there isn't really any VRM results out there to look at for the 7950/7970

[Ket]

Nobody keeps track of their VRM temps?

[Kain665]

Rule of thumb is 120c max for these chips. I personally don't like them getting to 100c. If those temps are the max you see then you're fine

[Andi64]

90'C with the stock XFX cooler. Room temperature is 30'C, or so. I've changed the i7 2600K+Z68-UD4 for an i7 3820+X79-UD3 and the case temperature is a little bit higher also.

[Ket]

Good to know all is healthy now guys 100c+ was WAY too hot for my liking as I frequently smelt a very toasty PCB possibly mixed with a little slightly melted solder smell but all that has gone since I redesigned the VRM cooling. I haven't measured VRM temps with my temp gun yet but they could be as low as around 60c as GPU-Z read my VRM temps a good 20c higher than what my temp gun reported prior to the redesign.

[Andi64]

I've decided not to buy the Arctic Cooling 7970 for my XFX HD7950. Instead, I bought a H60 bracket, with a 92mm fan over the VRM.

This one to be more specific: http://triptcc.com/product/h6080100-fan-bracket/

The i7 will get an updated H100i and I was planning on using the H60 on the HD7950

[Ket]

I considered something similar, but I decided it would use too much space which is why I picked up a Gelid Icy Vision A cooler. Max OC included (1155 / 1600), max GPU temp is 58c I think I'm going to get a 7w/km thermal pad though to knock off that final 3-5c on the VRMs. Cooler they are the better

[Rattle]

Reference 7970 3gb stock cooler

1125/1575 1.175v set 1.109v loaded according to gpu-z
max core temp 69c
vrm temp 1 59c max
vrm temp 2 50c max
fan speed maxed at 54% using fan curve profile in MSI AB, card bought new in may 2012 original TM and pads, cooler never removed.

[s1nykuL]

MSI 7970 Lightning stock cooler

1920x1080 res
1200/1500 @ 1.170v (GPU-z)

Max core:58c
Max VRM:59c
Fanspeed 50%

Hard to say if your temps are good or bad, The components on different cards by different manufacturers will not all be the same. So tolerances and maximum working temperatures would also differ. It would be useful if those with the same card posted their findings.

[Ket]

The VRMs Sapphire are using aren't what I would call normal, their very small and compact to other VRMs I've seen before. Heres a pic of them;



Difficult to tell if these VRMs are good or bad.. generally I'd say bad as hotter running components usually equal cheaper. That said the VRMs on my TwinFrozr hit about 83c in Heaven I think and the VRMs hat were on that card were what I would call standard.. so maybe its GPU-Z getting the temps wrong, my temp gun has already shown readings in GPU-Z are off.

[s1nykuL]

so maybe its GPU-Z getting the temps wrong, my temp gun has already shown readings in GPU-Z are off.

If that's so, then providing the same version of GPU-z is used, the errors should be consistent across the same family of GPU.
So if it is wrong for you it will be wrong by the same amount for others with the same GPU.

Providing GPU-z is over estimating you VRM temps I do not think 76c is a worry. Pushing 90c would worry me.

[King4x4]

Under an EK waterblock my gigabyte 7950s are currently 50-55'C with the cards clocked to 1200mhz.

[Matlad]

Ket, I have the same card, except mine is a 7970. I am overclocked to 1150/1650, my gpu temps after running heaven is 62-64 degrees and vrms are in the region of 58-62 degrees. I does sound like a Hoover however! I have ordered a prolimatech mk-26 which should get delivered tomorrow to quieten it down, will let you know the results. I think I will keep the sapphire vrm sink attached as it works well and just change the heatsink and get some better quieter fans on.

[Ket]

If that's so, then providing the same version of GPU-z is used, the errors should be consistent across the same family of GPU.
So if it is wrong for you it will be wrong by the same amount for others with the same GPU.

Providing GPU-z is over estimating you VRM temps I do not think 76c is a worry. Pushing 90c would worry me.

You would think so, but Sapphires own Trixx utility doesn't even work properly with the 7950 Vapor-X at this point in time. Software heavily relies on lots of different factors from components used on the PCB to the way the vBIOS is programmed so I certainly wouldn't rule out GPU-Z getting VRM temps wrong. I'll look into this further when the last few components arrive for the VRM cooling upgrade.

[Ket]

Ok, I've just finished upgrading the VRM and memory cooling, Heaven 3.0 has been running for some 15 mins+ now so I have most the results needed. Ambient temp 21.7c.

Heaven 3.0 Settings:

Extreme Tessellation
Shaders @ High
16x AF
8x AA
Catalyst AI set to "Very High"

Stock Cooling @ 55% (1155 / 1600):

GPU: 81c+
VRM1: 126c(!)+
VRM2: 85c+

What can we learn here? Well, that "upgraded-from-the-factory" cooling isn't worth the ink its printed in on the box. Secondly, sadly, I have no reading from my temp gun for this part of the test just GPU-Z 0.6.6 readings as the system continuously locked up... not surprising with those heinous VRM temps, especially on VRM1. Of course, its impossible to tell just how hot VRM1 was actually getting but taking into account my findings below we can estimate in the region of 104c+. All-in-all, wholly underwhelming performance from the stock cooling, VRM temps... the less said about those the better.

Upgraded Cooling:

Gelid Icy Vision A @ 43%
Aluminium BGA RAMsinks
Custom heatsink for VRM1
Re-designed VRM2 heatsink
Phobya 7w/km thermal pad (VRM2)
Arctic MX-4
Akasa 1w/km (best I could find) Thermal Tape (VRM1, vRAM)



GPU-Z Readings:

GPU: 56c
VRM1: 74c
VRM2: 71c

Temp Gun:

GPU: 61.2c
VRM1: 52.6c
VRM2: 72.5c

So, what have we learned? Firstly, GPU-Z 0.6.6 mis-reads GPU temps by 5c, a little skewed, but nothing drastic. VRM2 is read more or less bang on, but the surprise comes with VRM1, GPU-Z 0.6.6 mis-reads by a whopping 22.4c! Weather this is something specifically limited to just Sapphire cards I don't know, but all the same a temperature mis-read like that is concerning, I also expect temps to get slightly better for the VRMs as the thermal pad and thermal tape gets a little more cure / bonding time.

[Andi64]

I've installed the H60 on the HD7950.

Before the modifications my temps were:

GPU: 90'C
VRM1: 90'C
VRM2: 92'C

The XFX cooler really sucks, it's noise and it vibrates...

Now my temps are:

GPU: 51'C
VRM1: 60'C
VRM2: 62'C
(room temp is a little bit lower today, ~5'C or so)

The VRM right now doesn't have any heatsink attached, just a 92mm fan blowing air over it. I will attach some heatsinks later, to see if there is any difference. I also need heatsinks for the memory, I guess. Do you thinks it's necessary?

[Ket]

I don't think the memory really HAS to have heatsinks theres plenty of 7950s that don't and others just seem to have thermal pads slapped on them with a slab of metal on top, which I'd think hinders more than it does help. Still though, it is best to use heatsinks if you are aiming for the lowest possible temps.

[Andi64]

Testing Furmark:

GPU: 58?C
VRM1: 70?C
VRM2: 75?C

The GPU is really cool, but I need heatsinks on the mosfets! Anyway, much better than the stock cooler.

[Ket]

Those VRM temps seem about right if your getting the temps from GPU-Z. Since fixing the VRM cooling on this POS Vapor-X (well, POS until Sapphire at least release a vBIOS setting memory voltage to 1.6v, its set to 1.5v for some reason which just leads to crashing in games that use over 1GB of the onboard vRAM) my VRM temps appear to max out around 77c (GPU-Z) while playing very demanding games. Actual temps are probably more along the lines of 55c for VRM1 and 67c or so for VRM2.