Stren's Titan Water Block Roundup

[cr1t1cal]

Stren, when you gonna test Heatkiller waterblock?

[stren]

Stren, when you gonna test Heatkiller waterblock?

It was supposed to definitely ship on wednesday, I pinged them again to see if it did in fact ship

EDIT: It did in fact ship so hopefully numbers by the weekend

[cpachris]

So far I have always added thermal pads to the resistors regardless of the installation instructions. Should I continue this method or follow the manufacturer's instructions. I.E. Do what I think is best, or punish those manufacturer's who aren't as smart as me (jk jk)?

Both.

All you time and temperature data are belong to us!

[leopr]

Amazing work, i guess the safest bet is the XSPC Razor TITAN/780 + Backplate right ? I was about to pick two for my 2-Way SLI GTX 780 but now that i see the EK results i'm not really sure .

[stren]

Amazing work, i guess the safest bet is the XSPC Razor TITAN/780 + Backplate right ? I was about to pick two for my 2-Way SLI GTX 780 but now that i see the EK results i'm not really sure .

Yeah it all depends if you want better core temps or VRM/VRAM temps. The XSPC is a good choice, bear in mind my testing was done without backplate. Presumably the backplate helps the VRAM temps. You could even get some extra thermal pads and hit up the back of the VRM section too for extra cooling there.

[leopr]

Yeah it all depends if you want better core temps or VRM/VRAM temps. The XSPC is a good choice, bear in mind my testing was done without backplate. Presumably the backplate helps the VRAM temps. You could even get some extra thermal pads and hit up the back of the VRM section too for extra cooling there.

Thank you for answering, I guess the priority is the VRM cooling because i won't push the memory past 7000Mhz, do you think it is possible to re use the default thermal pads to cool the back of the VRM section ?.

[Vetalar]

it's a bit OT here, but i like idea to meet titan and chiller. they claim up to -50*C on core. or this is dedicated SS?

[techguy2k5]

I'm using the XSPC Titan block on my 780 and the results are fantastic! Full load (Furmark, Heaven 4.0, 3dmark 11, anything I can throw at it) never budges from 38 degrees C with boost clocks as high as 1293/6608 and 1.212V.

[stren]

I'm using the XSPC Titan block on my 780 and the results are fantastic! Full load (Furmark, Heaven 4.0, 3dmark 11, anything I can throw at it) never budges from 38 degrees C with boost clocks as high as 1293/6608 and 1.212V.

1293! Can we trade cards???

[stren]

it's a bit OT here, but i like idea to meet titan and chiller. they claim up to -50*C on core. or this is dedicated SS?

Yeah that's a dedicated SS not a chiller I believe Pretty sweet though for benching

[tiborrr]

Well, Titan is fairly easy to manage, heat wise. There's a rotary compressor in there, but at -50?C it will pretty much run in vacuum which will essentially kill the compressor. Plus all that cold on the memory will most likely coldbug the hell out of it. You will loose more with cold on the Vmems than you will gain with the cold on the GPU. This is why we normally run GPU only direct-die evaporators.

Better off building one by yourself or picking up pre-made unit by a renown builder.

[Mobsters]

Would it be a bad to add card original air cooler results too? just to see how much better water cooling is

[stren]

Would it be a bad to add card original air cooler results too? just to see how much better water cooling is

The problem is you're comparing core to ambient vs core to coolant. In an earlier post I said "I took some baseline measurements on air, with the fan at max (85%) the card was running about 50C over ambient which is not bad at all even if the fan is super noisy. " So you have to add on some amount to the liquid cooled results depending on how good your radiator/fan setup is at cooling vs the heat load you have. However most people wouldn't run at max fan rpm on the air cooler because it really is loud. So let's say your best air cooling is 75C with a 25C ambient, but your best water cooling with say a 7C gpu performance + a 3C delta for the loop would be 35C with a 25C ambient. On the other hand your worst water cooling loop might be 20C worse and your worst air cooling would be a lot worse of course as the card would overheat if you turned the fan too low. It's impossible to do a real apples to apples but those are your boundaries of best and worst case.

Does that make sense?

[Mobsters]

Does it make sense? yes and no. I wasnt really looking apples to apples, more like a baseline. Here is air cooler, here are water blocks.

What i am interested is vrm and vram temps mainly. Again this brings point what speed the fan is running etc. Lets just go with full fan 85% for best case and forget worst case as its just for start point

We all know that titan will slow down when core temp hits limit or power limit is reach. So core temp will be 80c or below in stock card depending on load as nvidia wants to keep things silent and people happy. So you could just add vrm and vram temps. There is no point adding core temps.

Interesting note, running fan at 85% will increase gpu power usage by 8% when compared to 30% fan speed. So theres 8% more power to overclock when in water

Awesome job with these results anyway

[Razor Time]

I honestly think it's more to do with contact on the resistors next to the VRMs and not all blocks do make good contact with those. I also think EK uses unusually thin thermal pads which explains it's results below. For example the swiftech didn't make contact with the resistors even though I added extra thermal pads on those resistors. Which leads me to the next question? My new probes came in so I can measure and log tons of points on the back of the board now. So far I have always added thermal pads to the resistors regardless of the installation instructions. Should I continue this method or follow the manufacturer's instructions. I.E. Do what I think is best, or punish those manufacturer's who aren't as smart as me (jk jk)?

Here's EK's VRM and VRAM temps. Very surprising.

Damn, I didn't realise you had replied. I think you could be right about the contact with the resistors as I have looked at the install pics for each of these blocks and its the only thing that makes sense. Your new results for the EK block look very impressive and it has kinda thrown a spanner in the works for me as I have already bought the Aqua block. Even though I have the Aqua block I haven't installed it yet so I could still return it if I so wish, but time is running out. I went for the Aqua block over the XSPC block as I wasn't totally convinced with the XSPC block after reading the other review which pegged it last, add to this I have always wanted to try out Aqua.

You mention that you placed thermal pads to the resistors regardless of the instructions. Did you place any pads on the Aqua block as I see they are the only block which doesn't use as much thermal pads compared to EK and XSPC? If you didn't place any pads here do you have any plans to do so in a future test?

You mention that you have a new probe, was the new EK reading done using this probe?

So if you were going to buy a block today for your own use which one would it be?

Sorry for the many questions but I want to make sure I buy and fit the best block I can.

[stren]

Damn, I didn't realise you had replied. I think you could be right about the contact with the resistors as I have looked at the install pics for each of these blocks and its the only thing that makes sense. Your new results for the EK block look very impressive and it has kinda thrown a spanner in the world for me as I have already bought the Aqua block. Even though I have the Aqua block I haven't installed it yet so I could still return it if I so wish, but time is running out. I went for the Aqua block over the XSPC block as I wasn't totally convinced with the XSPC block after reading the other review which pegged it last, add to this I have always wanted to try out Aqua.

Well the aquacomputer is still a good block, sure the VRMs are on the higher end, but VRMS are also generally expected to run hotter. So while cooler is better, I'm not sure how much additional life you would generally expect to get because it's very design dependent and time to failure on ICs while generally decreasing exponential with temperature, still has some average lifetime that is very design dependent. I design ICs myself so I can design a chip that can live at 110C for a long time, or I can design one that I know will die at 70C after a year. I suppose we might be able to find a datasheet if we really wanted to dig into that, but not many companies even give that data.

You mention that you placed thermal pads to the resistors regardless of the instructions. Did you place any pads on the Aqua block as I see they are the only block which doesn't use as much thermal pads compared to EK and XSPC? If you didn't place any pads here do you have any plans to do so in a future test?

Yes, every block so far has had them so that it was a "fair" test. If that is indeed fair lol.

You mention that you have a new probe, was the new EK reading done using this probe?

I haven't started using them. I didn't do any testing this week because I thought the watercool block would be here any day now as it shipped 12 days ago. Stupid DHL. Anyway I wanted to run the watercool with the same setup before I glue a mass of good thermal probes to the back of the board.

So if you were going to buy a block today for your own use which one would it be?

Honestly it's hard to say. Core temps are so close that it's tempting to ignore and choose on VRM temps and aesthetics. However if I were running 3 in parallel I probably would avoid the EK and run the XSPC due to EK's poor core temps at low flow. Not many of us can afford three titans though lol. I'd wait on Koolance until I can do a 2nd run to confirm good performance without warping the card. I also don't like just how hot the VRMs are on the Swiftech/EVGA block. On the other hand if I were choosing purely on aesthetics I think I'd buy the aquacomputer nickel plexi to be honest. Bear in mind we still have four blocks still to arrive though so everything could change yet...

Sorry for the many questions but I want to make sure I buy and fit the best block I can.

There will usually be a trade off in any decision and you have to weigh up what's best for you and there may not be a definitive block that sweeps every performance category. All of them will be an improvement over the stock air cooler. Even the swiftech VRM temps are still lower than the air cooler

[Razor Time]

Stren

Thank you for the detailed reply, its much appreciated.

When you say you used thermal pads for all the blocks I am assuming you placed extra pads as shown here in red on the Aqua block or have I miss understood you? If so what thickness of pads did you use? Do you think the temps would of been higher if you didn't place the pads here?




One thing I have noticed is that Aqua is the only block not to provide pads for this area and from memory every other gpu block I have used has always had pads in this area.

With regards to VRM temps I am not really concerned about the shelf life of the card but more of limiting my overclocking headroom if the VRMS are running hot. If this has no real effect on my overclocking headroom then I am not too concerned if the VRMs run 10-20c hotter than another block as they should be allot cooler than what they would be if I was running stock air cooling. It has been this thought that has been running constantly through my mind and one reason why I haven't added the block to my card. I actually have a GTX 780 and I want to get the best possible overclock out of it.

Thanks in advance.

[stren]

Stren

Thank you for the detailed reply, its much appreciated.

When you say you used thermal pads for all the blocks I am assuming you placed extra pads as shown here in red on the Aqua block or have I miss understood you? If so what thickness of pads did you use? Do you think the temps would of been higher if you didn't place the pads here?




One thing I have noticed is that Aqua is the only block not to provide pads for this area and from memory every other gpu block I have used has always had pads in this area.

With regards to VRM temps I am not really concerned about the shelf life of the card but more of limiting my overclocking headroom if the VRMS are running hot. If this has no real effect on my overclocking headroom then I am not too concerned if the VRMs run 10-20c hotter than another block as they should be allot cooler than what they would be if I was running stock air cooling. It has been this thought that has been running constantly through my mind and one reason why I haven't added the block to my card. I actually have a GTX 780 and I want to get the best possible overclock out of it.

Thanks in advance.

Yes that's correct.

I think it does affect overclocking headroom because I believe that the resistor is used to sense current and doesn't have zero tempco. I noticed a 4-5% change in the reported "power usage" when I didn't fit the thermal pads there even though the load was the same. Of course this doesn't matter if you have modified bios. In games I never saw any throttling with a 145% bios, but then my card is not capable of meaty overclocks anyway.

The required thickness of thermal pad varies from block to block. Generally I used the spares from that manufacturer's kit. The only exception was the hydrocopper where there weren't extras supplied. For the hydrocopper I used the koolance 0.7mm pads I think, but really I needed something thicker (or a double stack) as these didn't make a good connection to the block.

[Razor Time]

Stren,

That's interesting and partly what I have been concerned about. 4-5% is fairly high when there isn't much to play with in the first place. I hadn't planned on using a custom bios for my card as I can get over 1200mhz on air with a small voltage bump, but I may change my mind once the card is blocked. I am starting to think I might of been better off with a different block as decent overclocks are one of my main goals and I am now having second thoughts about my choice of block.

Can I ask if you used thermal pads or paste for the Aqua block memory chips?


Edit: After doing some reading it seems there are few reports of vrm temps effecting overall clock speeds on Titans. Very good info to know and thank you for pointing that out.

[Shoggy]

Can I ask if you used thermal pads or paste for the Aqua block memory chips?

Just in case you would use thermal pads on the RAM there is a good chance that the GPU will have no more proper contact to the block.

[Razor Time]

Just in case you would use thermal pads on the RAM there is a good chance that the GPU will have no more proper contact to the block.

Hi Shoggy,

The reason why I was asking Stren if he used thermal pads on the memory chips is that I am still trying to get my head around why the vrm temps are a fair bit higher compared to EK and XSPC considering all these blocks have a similar setup and all have direct contact. If thermal pads were used on the memory chips then this might possibly explain the difference in vrm temps???

Stren has already said he used extra thermal pads as per the picture I posted above which I know is not what you the manufacturer recommends in your install instructions. Would adding extra thermal pads to red circled area (resistors) effect anything? Stren has already mentioned that by adding thermal pads to this area he saw a reduction of temps on the vrms and less power usage on the card which sounds like without the thermal pads direct contact is not being made in this area. I am assuming the block was designed to make direct contact to this area?

Basically shoggy I love your block but it seems that vrm temps are fairly important when it comes to overclocking the GK110 chip which is what is concerning me.

[Shoggy]

The block was designed to have no contact to these inductors. I can not tell how much space is left between them and the hollow in the block but it should not be that much. Of course with an additional pad (which is too thick) it might be possible that the other parts in this area will have a lower contact pressure. Since the voltage regulators are next to the inductors it could have a negative effect.

[Razor Time]

The block was designed to have no contact to these inductors. I can not tell how much space is left between them and the hollow in the block but it should not be that much. Of course with an additional pad (which is too thick) it might be possible that the other parts in this area will have a lower contact pressure. Since the voltage regulators are next to the inductors it could have a negative effect.

Thank you so much for the reply Shoggy, it is much appreciated and helps explains Strens results.

[stren]

Stren,

That's interesting and partly what I have been concerned about. 4-5% is fairly high when there isn't much to play with in the first place. I hadn't planned on using a custom bios for my card as I can get over 1200mhz on air with a small voltage bump, but I may change my mind once the card is blocked. I am starting to think I might of been better off with a different block as decent overclocks are one of my main goals and I am now having second thoughts about my choice of block.

Can I ask if you used thermal pads or paste for the Aqua block memory chips?


Edit: After doing some reading it seems there are few reports of vrm temps effecting overall clock speeds on Titans. Very good info to know and thank you for pointing that out.

I did not use thermal pads only some paste on the aqua block wrt the memory chips. I could have used a bit more TIM on the memory chips to be honest as I would have thought the aqua block should have the best VRAM temps. On the retests I'll make sure to add a bit more.

Just in case you would use thermal pads on the RAM there is a good chance that the GPU will have no more proper contact to the block.

This is correct. I follow manufacturer's guidelines exactly on these.

The block was designed to have no contact to these inductors. I can not tell how much space is left between them and the hollow in the block but it should not be that much. Of course with an additional pad (which is too thick) it might be possible that the other parts in this area will have a lower contact pressure. Since the voltage regulators are next to the inductors it could have a negative effect.

Yup - if you add too much you're going to screw the VRMs. I would suggest some testing of the pad impression after fitting e.g. trial and error. I'll try and take photos of my pads after fitting on the next go through! The swiftech block may improve with thicker thermal pads for example. This is why I wonder if I should just leave the blocks as the manufacturer intended in which case the winners will be whoever said to put thermal pads on those components.

[Razor Time]

I did not use thermal pads only some paste on the aqua block wrt the memory chips. I could have used a bit more TIM on the memory chips to be honest as I would have thought the aqua block should have the best VRAM temps. On the retests I'll make sure to add a bit more.

Thank you for confirming Stren. Your time on this is much appreciated and has been very helpful. Keep up the good work.