Stren's Titan Water Block Roundup

[Church]

stren: i guess that precut/predrilled is only block in 660's kit, one has to cut/drill bottom plates by himself in 40/60 DIY miniblock bases. Imho it would be nice if for such product type community started sharing needed cut/drill dimensions for different gpu-s. Here a bit more offerings related to that, including spare baseplates.

[stren]

So as Alphacool, Bitspower and Phobya are being super super super slow I decided to change up the testing by adding actual data logged temperature sensors to the back of the card. Previously I was measuring the VRM/VRAM temps with an IR sensor gun. It was inaccurate by nature and prone to error on my part. So I bought some more probes and taped them down, used some MX2 to get a good contact to two spots on the board under the VRMs and then two more on two of the VRAMS:



Yes my workbench is a just a teensy bit untidy.



Then I applied liquid tape to secure the probes. It should isolate the probe a bit from the ambient air temperatures too hopefully. Wet liquid tape:



Dry liquid tape:





There she is in action:



The idea is to do three runs on each of the blocks that I have. With good time management I can do one run per day (each run has 8 hours worth data logging if I'm on top of it). Given that I have two VRM probes and two VRAM probes, I'm not sure whether to do a maximum VRAM or VRM temperature, or average the two probes. Either way the VRM/VRAM data should be more accurate than before, even if it's still not measuring the actual chips themselves.

Earlier I had been a bit flip-floppy in methodology for thermal pads and extra TIM and stuff. Koolance didn't like the extra TIM I had added, and it turns out that that was causing the bowing issues on the card. So a big apology to them. So all the testing from here on out will be done by the letter of the instructions even if I think I can improve it However despite this I'm not expecting data to be particularly different from the earlier runs. We'll see though! I should be done with all three Koolance runs tonight. Then because of all the data doubters I'll be running the XSPC next

I'm not sure how much value all of this work really is over what I've already done, but there's something to be said for doing the testing as well as you can even if it's another 4-5 weeks of work lol.

[Vetalar]

IMHO probes should be insulated from ambient air...

[stren]

IMHO probes should be insulated from ambient air...

While I agree, I think the liquid tape will do a "good enough" job, it's not like I'm measuring the VRAM on the cooled side, or the VRM body itself. This testing also misrepresents those blocks that come with backplates so it's far from perfect. At some point I'd like to get an accurate temperature logger that can use thin probes to slide in between the block and the VRM/VRAMs. The previous more inaccurate method was enough to see differences between blocks, and this should further clarify the difference.

Here's the data on the Koolance block - first round test vs the new average. Very similar which is good.



And the data from the new probes:



Much less noisy than the old method. There are significant differences between VRAM modules, primarily I think because furmark uses so little memory. VRAM 1 I had seen to be consistently the hottest when I would use my IR Probe. VRAM2 was selected as just another VRAM being on the far side from VRAM1.

[tiborrr]

EK SE (second edition) data:

The finer fins in the cooling engine did not alter the poorer performance at low flow.

The cooling engine is exactly the same (EK-FC Titan vs. EK-FC Titan SE).

[stren]

The cooling engine is exactly the same (EK-FC Titan vs. EK-FC Titan SE).

Then your press release should be edited better

About EK-FC Titan SE:
EK-FC Titan SE is a high performance full-cover water block for nVidia reference design GeForce GTX Titan as well as -GTX 780 series graphics cards. This water block directly cools the GPU, RAM as well as VRM (voltage regulation module) as water flows directly over these critical areas thus allowing the graphics card and it's VRM to remain stable under high overclocks. EK-FC Titan water block also features a very high flow design therefore it can be easily used in liquid cooling systems using weaker water pumps. The cooling engine now uses refined, more narrow microchannel structure to even further improve the heat transfer.

Good to know the results are similar then lol

[tiborrr]

Narrower (0.8mm) compared to older generation (1.0mm; GeForce GTX 600 series, Radeon HD 7000 series).

[stren]

So it seems like this might be all the data I'll get from this card. I'm getting artifacts now on any display output after the last two mounts I've done (total is now at 20 mounts or so). It's frustrating because a combo of precision-x not logging one run, and my usb drive dying meant that I lost data on 2/4 XSPC runs, and the third XSPC run would immediately crash under load so I had to discard the partial data from that run. So of the last 4 runs I only got one good set of data. That run does at least answer the biggest question thrown up by the first set which was is my data on the XSPC block repeatable. It still beats the Koolance on the new testing runs and is even better than the original run so I stand by the original data that suggested XSPC was the best on the core. Sadly we won't get the extra detail of averaged temps and better VRM and VRAM temps. So I'm going to call this one a day and write up the review as best I can with what I have.

edit -> spoke too soon, seems like it's actually the monitor that's artifacting and not the card... which I guess is good news. The monitor is 7 years old now so I guess these things are expected.

[Callsign_Vega]

Wow, glad I went with EK blocks as I am running the 1.3v softmod on my Titan's now which really stress the VRM's. EK destroys other cards cooling the power area. Is the EVGA/Swiftech block even contacting the VRM's? lol

[stren]

Wow, glad I went with EK blocks as I am running the 1.3v softmod on my Titan's now which really stress the VRM's. EK destroys other cards cooling the power area. Is the EVGA/Swiftech block even contacting the VRM's? lol

The EK does so well (I think) because it includes thermal pads for the inductors next to the VRMs which do get hot too. The thermal pads are also very thin. I agree though the VRM could be cooled better on quite a few cards and given that Nvidia have yet again given us such a weak VRM design for the power that serious overclockers inevitably try to run (given that no lightnings/classies were allowed for the hardcore). During this test I'm only running 123% power, guys like you running 1.3V are running much higher power from what I've heard (170% was a number I heard thrown around).

[Callsign_Vega]

Ah, only the EK block contacts the inductors? Good to know. I guess the other block manufacturers have to step up their game.

[stren]

Ah, only the EK block contacts the inductors? Good to know. I guess the other block manufacturers have to step up their game.

Well for those who don't have EK, if you're conservative with your thermal pads you can fit some spare pieces on them still. The only exception that I could see is the swiftech block where the cutout for the inductors is super high so you'd need to buy *thick* pads. I noticed that the air cooler has a cutout above the inductor so that the fan will actually blow some air over the inductors even though they aren't attached to a heatsink.

After the 20+ mounts I've done my card is now crashing after a little while of being under load regardless of waterblock and clock frequency that I try.

[Callsign_Vega]

That stinks, maybe one of the traces or something got damaged. Well thanks for the great roundup, hopefully it doesn't cost you a Titan in the end.

[theseeker]

Thanks for removing the guesswork Stren!

[m_jones_]

Thank you for all of this hard work, really hope your titan hasn't died.

[musicfan]

Stren, thank you so much for the countless hours of donated time and effort. So sorry, your Titan seems to be dying. It has been sacrificed for a good cause because you have taught us much. We are still digesting the data.

For XS members who would like to help Stren replace his Titan, we can donate to Stren's efforts at his site, particularly since EVGA never came through.

[stren]

Stren, thank you so much for the countless hours of donated time and effort. So sorry, your Titan seems to be dying. It has been sacrificed for a good cause because you have taught us much. We are still digesting the data.

For XS members who would like to help Stren replace his Titan, we can donate to Stren's efforts at his site, particularly since EVGA never came through.

Thanks musicfan - don't donate for a replacement yet though (I'll refund your donation). Although EVGA didn't come through with the loaner they may still come through with a warranty replacement. Of course this testing work might count as too much abuse to be under warranty also. But yes donations always help to fund the site, testing equipment and keeping everything advert free

[stren]

Alright I know some of you had requested pics of the internals on the cards:

AquaComputer






EK

First Edition:

Somewhat stained by the EK red coolant:



I had to up the contrast hugely to try and show the detail on the plexi block - still hard to see:



Second Edition:






Koolance:



Huge surface area there on both the main core and the VRM area. Seems like some staining that reminds me of the EK nickel flake issue going on...



Surprised given all of cooling surface area that it didn't top the charts!


Watercool Heatkiller GPU-X






Swiftech Komodo/EVGA Hydrocopper






XSPC

First remove the two layer aesthetic cover:



Internals look tiny now:



Remove the bridge connector

[Martinm210]

Awesome work as usual, sorry to hear the card troubles. After Skinnee and now you having cards fail, I don't think I will ever try a full cover VGA round. I have even managed to kill a couple of motherboards myself too. Hardware death seems to go hand in hand with significant remounting and VGA seems especially prone to it.

Anyhow, glad to see the data. I can't justify a titan let alone a waterblock, but it's nice to see the competitive efforts and relatively close results indicating all have been busy with block improvements.

Thanks!
Martin

[stren]

Awesome work as usual, sorry to hear the card troubles. After Skinnee and now you having cards fail, I don't think I will ever try a full cover VGA round. I have even managed to kill a couple of motherboards myself too. Hardware death seems to go hand in hand with significant remounting and VGA seems especially prone to it.

Anyhow, glad to see the data. I can't justify a titan let alone a waterblock, but it's nice to see the competitive efforts and relatively close results indicating all have been busy with block improvements.

Thanks!
Martin

Thanks Martin - turns out the card no longer crashes once I revert to the stock bios. I had to flash a new bios because the stock bios throttles so much as standard, but that throttling means that which ever solder joint is flakey never gets hot enough to break anything. The other possibility is that the bios somehow got corrupted so I can try flashing back to a "good" bios again. Bottom line is the card is alive for now

Finally got the full write up and video done:

write up: http://www.xtremerigs.net/reviews/wa...p-coming-soon/

video: http://www.youtube.com/watch?v=sDyYliSo-Hw

[m_jones_]

Thanks as always for this round up, it's a joke how Swiftech feel they can charge $150 for that block.

[leghmoh]

Stren, thank you very much for this review. You are the best

[musicfan]

Thanks stren. The text presentation is excellent as usual but the adding the video presentation hits it out the park. That's the best hour of video technology, I ever watched. I now subscribe to your YouTube channel and recommend that to others. You are a natural in front of the lens.

[stren]

Thanks stren. The text video presentation is excellent as usual but the adding the video presentation hits it out the park. That's the best hour of video technology, I ever watched. I now subscribe to your YouTube channel and recommend that to others. You are a natural in front of the lens.

Glad you liked it, this was my first "more professional" video, and also my first one in a long time, so I was a bit stumbly and waffley. I wasn't sure how many people prefer video over the written stuff so it's good to get feedback that it's useful, I'll try and do more in future then

[stren]

Just an extra note - since my review sample shipped (back in april?) AquaComputer updated the design a few weeks later to be higher flow, so essentially my AquaComputer results are now out of date by 4 months... I would retest if I could trust that titan, as it is we'll have to move on and reconvene on the next GPU