EVGA GeForce GTX 970 ACX has misaligned GPU vs heatpipes

#4919921

fantaskarsef

2014-09-23 09:25

Hmm, looks like something went wrong there. I wouldn't be surprised if the performance suffered if all the 970 ACX models are like that.

#4919932

Hilbert Hagedoorn

Administrator

2014-09-23 09:37

Yeah we have seen EVGA statements like that in the recent past. Talking their way out of things. Laughable is all I can say.

#4919933

Corbus

2014-09-23 09:39

Could it be that they used the same cooler thats on the 980 to cut down costs? Seeing that the 970 doesnt emanate as much heat,they thought it will do ok ...either way, d**k move.Curious to see how it looks on the 980.

#4919935

TheDeeGee

2014-09-23 09:40

All you need is a reference design card and a Arctic Accelero Xtreme 🙂 They are the Noctuas of GPU Cooling.

#4919942

Corbus

2014-09-23 09:54

All you need is a reference design card and a Arctic Accelero Xtreme 🙂 They are the Noctuas of GPU Cooling.

Yea, i started buying gigabyte thinking they are the same basically with their windforce coolers but no way.Arctic cooling is better and also a lot quieter. Definitely my next gpu will be reference with Accelero on it.

#4919948

Sukovsky

2014-09-23 10:06

Wow that's a shameful display. Was leaning towards MSI anyways so going that route

#4919949

GeniusPr0

2014-09-23 10:07

Lmao eVGA -_- What a joke

#4919960

Robbo9999

2014-09-23 10:24

Official EVGA response. "The way the EVGA GTX 970 ACX heat sink was designed is based on the GTX 970 wattage plus an additional 40% cooling headroom on top of it. There are 3 heat pipes on the heatsink – 2 x 8mm major heat pipes to distribute the majority of the heat from the GPU to the heatsink, and a 3rd 6mm heatpipe is used as a supplement to the design to reduce another 2-3 degrees Celsius. Also we would like to mention that the cooler passed NVIDIA Greenlight specifications. Due to the GPU small die size, we intended for the GPU to contact two major heat pipes with direct touch to make the best heat dissipation without any other material in between. We all know the Maxwell GPU is an extremely power efficient GPU, our SC cooler was overbuilt for it and allowed us to provide cards with boost clocks at over 1300MHz. EVGA also has an “FTW” version for those users who want even higher clocks. Regarding fan noise, we understand that some have expressed concerns over the fan noise on the EVGA GTX 970 cards, this is not a fan noise issue but it is more of an aggressive fan curve set by the default BIOS. The fan curve can be easily adjusted in EVGA PrecisionX or any other overclocking software. Regardless, we have heard the concerns and will provide a BIOS update to reduce the fan noise during idle. Thanks, EVGA" In short they know it's a **** design and are trying to get out of it

Referring to the bit in bold, I thought the same as EVGA when I saw the pics (even before reading the EVGA statement that you posted). The GPU die is not big enough to span the entire width of all 3 heat pipes, if they had shifted the GPU core to the right to contact all 3 heatpipes then there would have been 2 'gaps' (the bits inbetween the heatpipes) in contact with the core, whereas how they have designed it they only have 1 'gap' which is more efficient (a greater area of the die covered by heatpipe vs covered by aluminium block). As they say, the 3rd pipe (thinner pipe) is just there as a supplement to draw a little more heat away from the aluminium block. So, I don't think this off-centre die placement is an error, it's just part of their intended design to maximise percentage of die area covered by heatpipe.

#4919963

TheDeeGee

2014-09-23 10:32

I just noticed that one of those heatpipes aren't even "complete" and doesn't extend all the way. Wow EVGA, wow.

Seems all fine to me. It's the 6 MM one that's a bit shorter because there is a small raised area.

#4919964

Hilbert Hagedoorn

Administrator

2014-09-23 10:33

So, I don't think this off-centre die placement is an error, it's just part of their intended design to maximise percentage of die area covered by heatpipe.

I doubt that. You don't design and purchase a brand new revision 2.0 ACX three heatpipe cooler solution and then use only 2 of them just because it is convenient or 'by GPU design'. If it was done right the pipes would have been narrower and closer to each other, just like the example photo of the ASUS Strix in this post. I mean THAT is how it is supposed to be.

#4919965

TheDeeGee

2014-09-23 10:33

Referring to the bit in bold, I thought the same as EVGA when I saw the pics (even before reading the EVGA statement that you posted). The GPU die is not big enough to span the entire width of all 3 heat pipes, if they had shifted the GPU core to the right to contact all 3 heatpipes then there would have been 2 'gaps' (the bits inbetween the heatpipes) in contact with the core, whereas how they have designed it they only have 1 'gap' which is more efficient (a greater area of the die covered by heatpipe vs covered by aluminium block). As they say, the 3rd pipe (thinner pipe) is just there as a supplement to draw a little more heat away from the aluminium block. So, I don't think this off-centre die placement is an error, it's just part of their intended design to maximise percentage of die area covered by heatpipe.

Someone who understands 🙂

#4919966

fantaskarsef

2014-09-23 10:34

This could explain why EVGA's ACX cooler performs more poorly and noisier in reviews versus other custom cooling solutions.

From Hilberts news post. If this is true, they did something wrong still. I don't know, I haven't read many 970 reviews, and surely not compared different 970 custom coolers, but if they really perform sub-par to others, it was either a misconception, or outwright faulty manufacturing.

#4919974

BLEH!

2014-09-23 10:42

Stock cooler's more than good enough, no?

#4919977

Ven0m

2014-09-23 10:55

Stock cooler's more than good enough, no?

The stock cooler got improved, and when you pair it with significantly decreased power consumption, you get really good results. So if you want the air to be exhausted from the case by GPU cooler, the stock one should do the job. However, if I was buying one, I'd probably got with something like ASUS Strix, which should be pretty much noiseless under low load, and quieter/cooler with higher load, assuming that you have reasonable case airflow. For SLI setups, the stock cooler might be really good, perhaps even better than some non-reference designs. For example, when I was using 280x - MSI Gaming edition, the cooler worked really great in single-card setup, but in CF, one card was sucking hot air from another, overheating rapidly.

#4919978

TheDeeGee

2014-09-23 10:55

From Hilberts news post. If this is true, they did something wrong still. I don't know, I haven't read many 970 reviews, and surely not compared different 970 custom coolers, but if they really perform sub-par to others, it was either a misconception, or outwright faulty manufacturing.

If it's noisy, it most of the time means the Fanblade Design sucks. The ones Arctic Cooling use are dead silent for example, even at 2000 RPM.

#4919979

BrainDedd

2014-09-23 10:59

For the record ... the third heatpipe isn't even functional as the one end is not even closed off.

#4919984

Corbus

2014-09-23 11:11

For the record ... the third heatpipe isn't even functional as the one end is not even closed off.

The only reason why the third heatpipe isn't functional is because it makes no contact with the chipset, it is "closed off" because if you look closer the heatsink is sculpted there where the heatpipe should end, probably thats where the power connectors will be or some other parts might stick out. Anyway,its a poor design thats for sure. End of story.

#4919991

Ven0m

2014-09-23 11:28

To defend it, if the fins are connected to the base, that heatpipe wouldn't transfer much energy to begin with. But if it's intentional, it should be closer, or the base should be made of copper. Some infrared photos with properly set reflectivity would tell if it really works. Hilbert? 😀

#4919998

Hilbert Hagedoorn

Administrator

2014-09-23 11:37

Some infrared photos with properly set reflectivity would tell if it really works. Hilbert? 😀

Oh as you know I would have loved to investigate that for you guys. However EVGA has ex-communicated (read banned) Guru3D.com after we critiqued Precision X 15. So that is pretty much a NIL possibility.

#4919999

southamptonfc

2014-09-23 11:39

Oops! Schoolboy error!