Advanced search

Message boards : Graphics cards (GPUs) : Raijintek Morpheus VGA Cooler

Author Message
Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2334
Credit: 16,178,080,749
RAC: 281,455
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41941 - Posted: 4 Oct 2015 | 18:45:53 UTC
Last modified: 4 Oct 2015 | 18:49:04 UTC

I've recently changed the standard cooler of my GTX 780Ti to this Raijintek Morpheus cooler, with two Scythe Kaze jyu 100mm 2000 rpm fans (SY1025SL12M), and I'm quite amazed that the card is running at 56°C instead of 75°C, with much less noise.
You can witness the change in this task's stderr.
Roused by this, I've changed the cooler of my standard GTX980 also, and it runs at 45°C instead of 70°C.
The cons of this cooler are:
1. It's very wide (especially with two standard 25mm thick fans), so it's not really applicable for multi-GPU systems
2. The aluminium heat sink pieces are not fully compatible with these cards, as I had to cut the VRM heatsink in half for the GTX 780Ti, and use the leftover memory heat sinks for the GTX 980, as the FETs of its VRM have a quite different arrangement, and the supplied VRM heatsink can't be applied to a GTX980. Moreover the RAM chips in the GTX980 are closer than on the GTX780Ti, so I had to cut the edges of the RAM heat sinks to avoid it to contact with the main (GPU) heat sink's heat pipes.
3. The supplied thermal twin adhesive didn't stick to the RAM chips (only to the heat sink), so I had to use a leftover thermal adhesive from Arctic Cooling Xtreme.

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2334
Credit: 16,178,080,749
RAC: 281,455
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41950 - Posted: 5 Oct 2015 | 20:28:37 UTC

I'd like to add an important notice for those, who change the original heatsink / cooler assembly on their GPU:
The metal frame, which acts as the memory / VRM FET heatsink, also acts as reinforcement, making the card rigid.
Without this frame, when the card is fitted horizontally its PCB could be easily warped by its own weight (plus the weight of the PCIe power cables), shortening the lifespan of every soldering, thus of the whole card.
Some factory OC-ed cards are manufactured without such frame, so it applies for those cards too.
To avoid warping the PCB, I underpin the inner end of the card (at the PCIe power connectors) with a proper sized stick (actually I've cut a bamboo skewer to the needed length). There are other (more professional) accessories for this purpose.

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 41953 - Posted: 6 Oct 2015 | 8:54:45 UTC

Impressive cooling performance! I do wonder though how such solutions improve cooling of other critical components, such as the VRMs and memory chips (esp the VRMs). Even with high quality solutions, such as EVGA, these components are treated as second-class citizens, left to be cooled by smaller heat sinks or metal plates and thermal pads. One can say, these cooling solutions are capable enough. I do not however feel very comfortable with components in my computer working 24/7 at 90-100C. Of course, others may feel absolutely comfortable!

With Maxwell, one could think that VRMs do not get so hot as with Kepler or Fermi, since it requires less power. Less power can also be delivered with fewer VRMs working harder however, so cards end up with fewer, but also equally (or almost) hot VRMs.
____________

Betting Slip
Send message
Joined: 5 Jan 09
Posts: 670
Credit: 2,498,095,550
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41954 - Posted: 6 Oct 2015 | 8:58:51 UTC - in response to Message 41953.
Last modified: 6 Oct 2015 | 9:01:10 UTC

If your cards are not HOT they aren't working hard enough :-)

However, GPUGrid has a better cooling solution, no work available. :-( Or have a lot of users got a stash?

Thomas H.V. DUPONT
Send message
Joined: 20 Jul 14
Posts: 732
Credit: 100,006,094
RAC: 0
Level
Cys
Scientific publications
watwatwatwatwatwatwatwat
Message 41955 - Posted: 6 Oct 2015 | 11:19:56 UTC - in response to Message 41941.

Thanks for the heads-up Zoltan !
Really interesting :)
____________
[CSF] Thomas H.V. Dupont
Founder of the team CRUNCHERS SANS FRONTIERES
www.crunchersansfrontieres.org

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2334
Credit: 16,178,080,749
RAC: 281,455
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 41959 - Posted: 6 Oct 2015 | 15:14:27 UTC - in response to Message 41953.
Last modified: 6 Oct 2015 | 15:17:48 UTC

I do wonder though how such solutions improve cooling of other critical components, such as the VRMs and memory chips (esp the VRMs).
It usually gets better too, because the metal frame these components are attached to has no fins, it simply spreads the heat across the whole card. Moreover the thermal pads between the components and the metal frame is relatively thick, compared to the GPU's thermal interface material, resulting in poor thermal conductivity.

But there are examples for VRM thermal increase:


Even with high quality solutions, such as EVGA, these components are treated as second-class citizens, left to be cooled by smaller heat sinks or metal plates and thermal pads.
Compared to the heat produced by the GPU they are really second-class :)
The VRM's efficiency is between 90-95% (later, or OC models usually have better FETs), so while the GPU emits 200W heat, the VRM emits 10-20W, and the RAM chips emit about the same amount, but they are spread across a much larger area than the GPU chip. The challenge of cooling them comes from that they are very close to a 10 times stronger heat source (the GPU).

One can say, these cooling solutions are capable enough. I do not however feel very comfortable with components in my computer working 24/7 at 90-100C. Of course, others may feel absolutely comfortable!
A FET in the VRM, or even the RAM chips are way simpler / smaller / cheaper than the GPU, so replacing them is not a big deal, and they aren't that delicate as the GPU. Still - you are right -, the lower their temperature the longer they will work. Besides that the high temperatures wear the chips more, the whole process of heating up and cooling down (aka the thermal cycle) wears the soldering and the layers of the PCB & the chip more when the thermal peak is higher. That's one aspect of the dangers of LN2 cooling: when you apply the LN2, the chip goes from room temperature to -196°C, which is 3.5 times more change as of going from room temperature to 80°C.

With Maxwell, one could think that VRMs do not get so hot as with Kepler or Fermi, since it requires less power. Less power can also be delivered with fewer VRMs working harder however, so cards end up with fewer, but also equally (or almost) hot VRMs.
That's why OC cards have more "phases" in their VRMs: less heat emitted by each component in a larger area.

Vagelis Giannadakis
Send message
Joined: 5 May 13
Posts: 187
Credit: 349,254,454
RAC: 0
Level
Asp
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 41972 - Posted: 8 Oct 2015 | 9:20:23 UTC - in response to Message 41959.

Thanks for your response Zoltan! You are absolutely right in your observations / conclusions.

Myself, I wasn't even aware of the other heat sources on a graphics card, until I read an article on Tom's Hardware on the thermal behavior of different implementations of a specific Nvidia GPU (I can't remember which). It was in that article that I saw how different manufacturers (top, middle or "cheap") treat those other heat sources in different versions (again top, middle or "cheap") of their cards: range is from total indifference ("let them cool themselves off") to dedicated parts of the card's cooler. I figured, for 24/7 demanding operation, cooling all the heat sources is important and not all cooling solutions are equal.
____________

Post to thread

Message boards : Graphics cards (GPUs) : Raijintek Morpheus VGA Cooler

//