Benchmark Testing
Gaming benchmark tests were run against the GTX 970 SC ACX 2.0 graphics card using the ACX 2.0 cooler and HeatKiller GPU-X3 GPU water block in stock and overclocked configurations. Results were provided to give a more complete understanding into the potential performance benefits gained from using the DirectCU H2O cooler in conjunction with an air-based or liquid-based cooling configuration. Note that only a single set of numbers are shown for the card's stock performance numbers because the +40MHz boost in GPU clock speed while liquid cooled was found to have neglible effects on the card performance.
FutureMark 3DMark
FutureMark Corporation’s 3DMark is one of the most grueling benchmark tests on the market for determining system graphics performance. The Fire Strike benchmark measures both graphics-based and PhysX-based system capabilities. The Fire Strike benchmark was run at default settings with a 1920×1080 resolution three times with the highest reproducible 3DMark scores recorded.
The Fire Strike benchmark tests demonstrate the expected scaling for the card. The ACX 2.0-base overclocking increasing performance by about 10% and the HeatKiller-based GPU cooling adding another 5% over the air-cooled overclocking performance. A 15% performance boost for liquid cooling is not too bad. The PhysX-related operations do not seem to scale with the card GPU or memory speed, remaining virtually constant throughout all runs.
Unigine Heaven 4.0
The Unigine Heaven 4.0 benchmark is one of the more strenuous methods to determine a video card's performance. The Heaven benchmark was run a 1920×1080 resolution with all graphics settings set to High, On, or Normal with the following exceptions: 8x Anti-aliasing and DX11 API selected. The benchmark tests were run three times with the highest reproducible FPS scores recorded.
The GTX 970 SC's performance continues to scale as expected with the liquid-based overclocking squeezing out just under a 15% gain over stock, while the air-cooled overclock managed just under 10%.
Batman: Arkham Origins
The Batman: Arkham Origins in-game benchmark was used to determine approximate real-world performance with the card under test. The benchmark test was run at a 1920×1080 resolution with all graphics settings set to High, On, or Normal with including the following settings: GeForce TXAA High Anti-aliasing, DX11 Enhanced mode enabled where applicable, and Hardware Accelerated PhysX High. The benchmark tests were run three times with the highest reproducible FPS scores recorded.
Again, the card's performance scales as expected with the liquid-cooled GPU able to squeeze out just under 15% more performance in comparison to the stock numbers.
Metro: Last Light
The Metro: Last Light in-game benchmark was used to determine approximate real-world performance with the card under test. The benchmark test was run at a 1920×1080 resolution with all graphics settings set to High, On, or Normal with the following exceptions: SSAA enabled, 16x Anisotropic Filtering, and Advanced PhysX disabled. The benchmark tests were run three times with the highest reproducible FPS scores recorded.
Metro: Last Light proves a bit more challenging to the GTX 970 SC performance-wise with the liquid-cooled card managing a performance gain of just over 10% in comparison to the stock numbers.
Grid 2
The Intel AVX-enhanced version of the Grid 2 in-game benchmark was used to determine approximate real-world performance with the card under test. The benchmark test was run at a 1920×1080 resolution with all graphics settings set to Ultra, High, On, or Normal with the following exceptions: 8x MSAA Multisampling. The benchmark tests were run three times with the highest reproducible FPS scores recorded.
Again, we see the card scaling as expected in Grid 2 with a cool 15% increase in performance over stock with the HeatKiller block cooling the GPU.
Why didn’t this come around
Why didn’t this come around when I was building a new rig so i wouldn’t have to get a 980GTX for a watetblock?
Now you have a 980 to try it
Now you have a 980 to try it out on?
I need to go back to reading
I need to go back to reading the articles, its a mod not a card. /derp
this just makes me wish all
this just makes me wish all cards came with an AIO liquid cooler option, hopefully AMD’s new offering will help make this a trend.
look at the performance of evga’s AIO gtx980ti, and that is a reference board.
nice write up and hope all is well.
IMO this think looks cheesey
IMO this think looks cheesey and cheap. those stick on ram coolers, yea umm no. who knows how long they will stay stuck on to the ram before falls off and who knows what will happen.
IMO if you are gonna water cool should just go with a full coverage block my evga hydrocopper.
I have a machine with an 8,
I have a machine with an 8, year old gpu in it using those ramsinks and not one has budged… I think you’re a bit too paranoid
I’ve had great experience
I’ve had great experience with Swiftech universal blocks. What I like about them is you can reuse them on any new card you go with, versus having to buy a new full coverage block every time you go to upgrade and you want to water cool.
They pretty much do make an
They pretty much do make an AIO version of the 970….But a AIO cooler and then buy the nzxt bracket to mount the cooler onto the card. It is SUPER easy to do and it dropped my temps by 20 deg
Personally, I didn’t see
Personally, I didn’t see enough of the upside to do this.
You saw the load temps,
You saw the load temps, right?
I’m a clumsy dumb ass who has
I’m a clumsy dumb ass who has bricked a card with an aftermarket cooler. I miss the days when a watercooler could almost double performance. So I can see folks holding off, but the noise reduction is phenomenal.
I watercooled both my 970
I watercooled both my 970 ftw’s and have them overclocked to over stock 980 specs. I could not accomplish this without the watercooling. I don’t think it doubled performance but it added a TON.
Personally, I’ll leave it up
Personally, I’ll leave it up to the overclocking audience. My issues are mostly general user. In short- my arguments are not for the challenge of overclocking as having moved on from that interest.
Overclocking on this kind of margin is generally overcome by the next generation of video cards. The heat won’t make a compelling difference in the warranty of the GPU if you don’t overclock. The heat issued doesn’t change very much- it’s only displaced somewhere else- notable for those who pay AC. But the thing I noticed in this one was the sound. I’d have to look at a fan as well. I have to admit I’ve come to love my mouse quiet system and video card. I game less than I work so a handful of FPS just isn’t worth it.
Well, enough of being a drug. I’ve decided that if I’m going to look into this effort, I want a notble bang.
PS, I have a 980- that might be fun too- if a little more pricey.
Mouse quiet? I don’t know
Mouse quiet? I don’t know about that! I have a little pet rodent and He is very loud sometimes! Especially if his bowl is empty or he wants picked up. And when I am sleeping he always bangs on his wheel until I let him out, Then after I let him out he keeps me awake from jumping up and down of the bed or walking on my head! So no! Little rodents are very noisy!
Well, of course they look
Well, of course they look impressive at first sight, but keep in mind: the air cooled temps are still solidly within nominal range with lots of reserve to spare, even though the cooling uses the silent profile.
And the low temps are bought with a much higher idle noise level. And yes, pcper, you can still go to Zero speed when idle, even when OC. Makes me wonder if they even tried to adjust the fan curve to achieve a stable OC with minimal noise or just went with the full blown Agressive setting (to make the watercooler look at least somewhat competitive?) – the far lower delta temps do suggest the latter.
And what for? For a ~3-3.5% increase from air to water OC, and an impressive, but ultimately pointless low delta T?
Sorry, not impressed.
Nice article, although I
Nice article, although I don’t think Maxwell benefits too much from improved cooling. I’d like to see the benefits on a 290X or something else whose overclocking is generally limited by temps rather than arbitrary power limits.
I have been looking into
I have been looking into doing something like this on just one of evga 970s, the top one gets very hot. Thanks for the write-up.
I have two Evga 970 SC models
I have two Evga 970 SC models just like the one used here. Despite the cards running relatively cool, I can’t get them past 1404Mhz reliably. They always seem fine (no artifacting and the temps are fine), but eventually after around 3hrs under load the NVidia driver fails. This doesn’t happen in all games, and hilariously it doesn’t happen in furmark. Is this a driver thing or are my cards just poor?
Also, I used to have stick on heatsinks on a modified X700 from years ago. I remember my memory overclock fell off a cliff one day and I opened my case to discover a few of the heatsinks sitting on the floor of the case…
Thanks for doing this test to
Thanks for doing this test to show us what is possible, but for the time and money that would go into something like this negates the positives imo. I already got a 1516MHz actual in-game boost clock using the stock air cooler on my GTX 970, and fan noise (auto) and temps (83C) are not an issue.
The stock fans on my H100i need to be swapped out for something quieter though…
I have had 6 of the black
I have had 6 of the black cougar vortex PWM fans for a few years and I can vouch for there quietness and good airflow. They direct the air strait out of the fan,not in a cone shape like most fans do. Their build quality is fantastic and they have a very very long unrealistic lifespan listed for them of 300,000 hours.
I had them on my 4.5-5.0 ghz 2600k’s original H-50 and they performed better then the enermax magma’s I had on it originally but in the last 5 months I upgraded the cooler to the fantastic 38mm thinck radiator Cooler Master Nepton 140XL and it came with 2 140mm jet flow fans that are crazy loud at 80-100% but at 60% and under thy become pretty silent and they have great static pressure and even with a ambient temp around 29c the 2600k at 4660mhz tops out at 67c after 20 minutes of Intel burntest v2 with the fans at 65%
Good review……but it
Good review……but it definitly needed some other cards benchmarked for comparison. Like a regular 980, 290/X etc,but the regular 980 is what I think people would like to see if a 150 or so dollar GPU can hang with a higher tiered card.
Also I realize you used a waterblock you had on hand but any real water cooling fan would use a full cover waterblock, they look better and perform better buy keeping the vrms and other hotspots near ambient temperature which is paramount when pushing a card at its highest function able overclock especially if you want it to survive longer than 6 months to a year.
Water cooling and modding is
Water cooling and modding is not just about putting things together with off the shelf parts, but experimenting with what you have to see if you can get decent performance without killing your budget. I've been water cooling my rig for a long time (over 10 years) with many iterations in my build and cooling apparatus, so I think I more than qualify as a "real" water cooling enthusiast. This was more of an experiment to see how well it would work, and it seemed to work pretty well in my opinion.
Thanks for the feedback though…
How far can you push that
How far can you push that card if you do the BIOS mod to increase the power limit, and voltage limit to 1.285V?
That’s exactly what I was
That’s exactly what I was thinking.
Or more importantly how far can you push a Ti or Titan with a custom Bios and water cooling?
At stock you can only increase the TPD by 10% on the Ti & Titan but you can increase by 25% on the 970 & 980. So opening up that extra power could provide a very good boost.
Epic fail that the VRM’s
Epic fail that the VRM’s aren’t watercooled, wrong water block chosen if you ask me.
I’ve been running 2x GTX 970
I’ve been running 2x GTX 970 in SLI on a mATX board in a mATX case hooked up to G20 AIO liquid cooler brackets on corsair 140mm kits. They run about 40C all day even heavily overclocked. Really good set up.
if they sold these complete
if they sold these complete I’d think about it
cant get my evga gtx sc past
cant get my evga gtx sc past on stock cooler, but is stable
GPU 1467
MEM 3764