[h] 4 Weeks with NVIDIA GeForce GTX 980 SLI

[96Firebird]

That's fine if that's how it is engineered. Something like the Asus DC cooler though is designed for the heat pipes to make Direct Contact with the GPU.

How do you expect them to ensure all the heatpipes touch the GPU? Shrink the heatpipes, and they become less effective. Add an intermediate plate, it is no longer direct contact. Looks to me like they have their largest pipes contacting the GPU. How would you engineer it so the other pipes touched as well?

 

[SolMiester]

deleted

 

[3DVagabond]

How do you expect them to ensure all the heatpipes touch the GPU? Shrink the heatpipes, and they become less effective. Add an intermediate plate, it is no longer direct contact. Looks to me like they have their largest pipes contacting the GPU. How would you engineer it so the other pipes touched as well?

They would need to do it the same way the AMD exclusive AIB partners do it by designing a cooler dedicated to cooling Hawaii. Coolers by Powercolor, Sapphire, and XFX don't have any issues keeping the GPU as cool as nVidia partners keep the GK110. It's when they reuse a cooler that was designed for the larger chip that issues occur. You can get away with it if the chip is using 25%-40% less power. That's why it's OK with GM104.

 

[Keysplayr]

Anyway, back to the topic. Kyle was most impressed with the power consumption and heat output of the 980s over 290Xs.
I'd love to see him throw in a pair of 970s and use them for a month and let us know what he thinks.

 

[wand3r3r]

Here's the rest in case you missed it.

http://www.hardocp.com/article/2014/10/30/4_weeks_nvidia_geforce_gtx_980_sli/4

The Bottom Line


Have I been happy with my Radeon R9 290X CrossFire system? Absolutely. A new NVIDIA GeForce GTX 980 SLI setup will easily remind you what you would like changed about it however. The GTX 980 cards are so much cooler, and so much quieter than the 290X cards, it will astound you that these cards perform so incredibly well. The efficiency that NVIDIA has accomplished with this new Maxwell GPU is praiseworthy.



From a pure gaming performance standpoint, I am not upgrading in my move to 980 SLI from 290X CrossFire. From a comfort and usability perspective, the 980 SLI is twice the product 290X CrossFire is, or maybe a better way to say it would be that 980 SLI is half the product that 290X CrossFire is...half the heat and half the noise. Sadly that thinking does not extend to the half the price.

Today you can purchase a custom cooled Radeon R9 290X card for as little as $320 after MIR. The least expensive GeForce GTX 980 card for sale today is $549. If you are taking SLI into consideration that is a fairly big price bump for better efficiency. You can argue electric costs over time, but the simple fact is that it would take a hell of a long time to recoup $460 no matter how many hours a day you gamed.



Even though I am specifically talking about my multi-GPU experience, single card GTX 980 performance cannot be ignored. When you use the NVIDIA GeForce GTX 980 in a single card configuration on a sub-4K single display, the GTX 980 not only is more efficient, it is simply a better performer, and by a good margin. Overclocking your GTX 980 will push the performance even further out from the R9 290X in performance.




For 290X CrossFire users, it is very hard to make the case that it is worth it to jump to a GTX 980 SLI setup. I am not writing this to try to sell you on one or the other. Even if you could sell your used 290X cards for $500, another $600 for a set of GTX 980 cards in order to get better efficiency, less heat, and less noise is a fairly steep cost.




If you are considering a new multi-GPU system, the pricing on the current crop of R9 290X cards makes CrossFire a lot more palatable wallet-wise than GTX 980 SLI. For a long term multi-GPU purchase, there is no doubt in my mind that you would easily be happier living with GTX 980 SLI day to day.

 

[Headfoot]

But to place the blame on heat pipes not contacting the GPU is silly.

Hardly.

Basic heat transfer physics: the more things the heat must pass through, the less efficiently heat moves because in the real world no transfer medium is perfect. Or was I imagining the reduced temperatures across the board in 05-07 when CPU coolers started going direct heatpipe?

At worst the outside pipes get almost no thermal load to distribute. At best, the outside pipes get less of a thermal load to distribute. There is NO scenario where the outside pipes conduct exactly as much heat as the direct touching pipes. That is a physical impossibility. However, it is a matter of degrees (literally).

The real question is "How much does no die contact by pipes 1 and 5 increase total thermal resistance of the cooler (i.e. reduce effectiveness)?" Some of you guys say "hardly at all," other guys say "quite a bit." The guys saying it makes a tangible difference have multiple reviews where the two full-contact coolers (PCS+ and Tri-X) have better temperatures by quite a margin. The guys who say "it makes no difference" have no reviews that show the MSI and ASUS DC coolers as equal or better. Nor can they provide evidence that the PCS+ or Tri X are superior by virtue of some other attribute that the MSI and ASUS DC lack. Both groups of guys rely on an inference. The "tangible difference" guys rely on an inference PLUS evidence congruent with their theory.

The burden is on the "No difference" guys to back up their hypothesis with evidence. Speculation that it's not because of the heatpipes contradicts the simplest explanation and requires evidence to substantiate that position. TLDR: citation needed.

 

[tolis626]

Although I tend to agree with the guys that say direct heatpipe contact makes a big difference in cooling, there is one thing no one has mentioned yet. Coolers designed for Hawaii all use 3 fans instead of 2. That may not account for all the difference, but it's bound to make at least a small amount of difference. Other than that, it's all about the efficiency of the underlying cooler design. Transfering heat directly from its source (The GPU) to all the heatpipes at once is more efficient than transferring it from the source to 3 out of 5 heatpipes, then let the metal in the heatpipes transfer it to the heatpipes that aren't in touch. If anything, these things get saturated with heat at some point, and efficiency takes a big hit at that point.

All that changes if all heatsinks are inside a single copper block, but not dramatically so. Restrictions still apply. One can't fight physics unfortunately. With all that said, I think we've made a very big deal over a not-so-big issue. Cards like the Tri-X and PCS+ provide better cooling, that's about it. Why it does so? I don't really care that much that I'll lose any sleep over it.

 

[Silverforce11]

Gigabyte WF3 has 3 fans and the first iteration with the 2 pipes without contact was horrid, you can check the reviews on computerbase.de and tomshardware(de) where they go specifically into this problem. The 2nd updated version has a copper heat plate and it was much improved but still not a great performer.

XFX DD for R290/X has 2 fans, it performs just fine.

 

[96Firebird]

Nor can they provide evidence that the PCS+ or Tri X are superior by virtue of some other attribute that the MSI and ASUS DC lack.

I already stated my hypothesis on why the PCS+ and Tri-X coolers are superior to the MSI and DC coolers. The heat pipes are always in contact with the heat sink, instead of traveling outside the card to join the heat sink later on.

Now the burden is on you to prove me wrong.

Edit - Which MSI card are you talking about? I can't find one that has heat pipes that directly contact the GPU...

 

[greybaby]

Good read, thanks for the link

 

[Headfoot]

I already stated my hypothesis on why the PCS+ and Tri-X coolers are superior to the MSI and DC coolers. The heat pipes are always in contact with the heat sink, instead of traveling outside the card to join the heat sink later on.

Now the burden is on you to prove me wrong.

Edit - Which MSI card are you talking about? I can't find one that has heat pipes that directly contact the GPU...

It doesn't work like that. You don't lay down speculation then tell everyone to disprove you. Provide evidence for your initial position, then we'll talk.

 

[3DVagabond]

It doesn't work like that. You don't lay down speculation then tell everyone to disprove you. Provide evidence for your initial position, then we'll talk.

The evidence is that the coolers work better.