Stren's R9-290/290x Water Block Testing

[stren]

This thread is awesome and so helpful thanks a ton!

You say the passive AC backplate works with the EK block? Did you the test Rev. 1 or Rev. 2 of the EK block? Reason I'm asking is my card (MSI Twin Frozr) only works with Rev. 2 of the EK block.

Also if the passive backplate works shouldn't the active one work as well since the screws holes are the same?

I tested rev 1, in the past I have tried the rev 2 and saw no real difference, usually it's more about resolving clearance issues. I think the MSI has glue on top of the inductors and so VRM results may get worse in general.

I believe that the active one is the passive one plus the heatpipe that connects to the bridge. The heatpipe will be compatible with the backplate, but the bridge which needs to connect to the block is extremely unlikely to be compatible with the EK block.

More great detailed info. Any thoughts on the possibility of closing the gap between the EK and AC backplates by just replicating the extra thermal padding of the AC plate on the EK plate?
The EK plate seems to make only a marginal improvement.

I do think it would close the gap but not sure how much - the thermal pads will most likely have to be thicker as the AC backplate comes closer to the PCB in those areas I think. Thicker is of course more thermally resistive, but you could counter this by buying better thermal pad quality to lower the resistance.

[JaseUK]

Anyone know if there is a backplate as good as the aquacomputer one as its very difficult to find anywhere that stocks them?

[techjesse]

Excellent review, I really enjoy reading your reviews Stren

[stren]

Anyone know if there is a backplate as good as the aquacomputer one as its very difficult to find anywhere that stocks them?

I haven't tried all of them (not all mfr's send them) however the AC is the only one I've seen that produces a significantly different result to the block on it's own.

Excellent review, I really enjoy reading your reviews Stren

Thanks!

Things have been delayed again after the Impact's battery died (again) and the clock would reset every hour during testing yielding bad data logs. I got a new battery and now after the bios lost it's settings furmark seems to be crashing during testing. I assume they are not related, but who knows. I'm really not in love with the impact. Lots of niggling and weird stuff that the latest bios still doesn't fix. That and Asus' denying warranty coverage on the first board has really put me in a bad mood towards it lol.

Anyway I need to try downclocking the card to see if it becomes stable but afterburner won't let me set the clock lower than 1150 so I may need to reflash the bios and go from there. If I have to restart testing again for the fourth time I might lose it.......

[stren]

So I managed to fix things "temporarily". I couldn't downclock the card from my overclocked settings even after a gpu bios reflash and a reinstall of afterburner. In the end updating drivers to 14.6 and then back to 14.4 fixed things. I was able to complete a day's run on the heatkiller block, before clicking delete data rather than export data :madface:. I powered down and the next day I went to go take data again (today) and the damn thing won't post or go to bios. The weird thing is that the motherboard claims no boot device, yet swapping the GPU seems to enable it to go to the bios as normal. I'm suspecting the GPU is dead and the driver reinstall was merely a reflection of it being close to death. For some reason it waited until the next power up to die. I have my fingers crossed that it's not dead. There's a chance that a reinstall of a block might fix something, but I'm not feeling lucky given everything else that's gone wrong (dead motherboard, psu and gpu #1).

[stren]

Figured it out - AX850 8 pin cable had fused as it came out of the PSU. Card looks to be fine despite whatever must have caused that, hopefully using a different PSU won't then kill it. I don't believe that PSU is under warranty as it was a cheaper refurb unit that I think only has 90 days of warranty. Hopefully testing will continue smoothly this week! Pic of the fused cable that can not be removed:



At least I didn't burnt the house down...

[stren]

Figured it out - AX850 8 pin cable had fused as it came out of the PSU. Card looks to be fine despite whatever must have caused that, hopefully using a different PSU won't then kill it. I don't believe that PSU is under warranty as it was a cheaper refurb unit that I think only has 90 days of warranty. Hopefully testing will continue smoothly this week! Pic of the fused cable that can not be removed:



At least I didn't burnt the house down...

Alright Stren has been busy in the background - retested a bunch and close to being done finallly! Couldn't get a good mount with the XSPC block so Paul shipped me a new one - gonna get a retest done on that today. Gabe also requested a retest because of the VRM temps and pics be put up of Swiftech's block so I'll get that done. Alphacool is also sending a block so we should have very good manufacturer coverage

[stren]

While we wait for final results to come through here are some pics of the blocks that I didn't show already:







Accessories including color gels to change the LED color:



Block now has a window and as usual with Swiftech the backplate is included:



You can see raised areas on the backplate to cool some IC's on the back of the PCB:



I do feel like the logo is still obnoxiously large on the backplate - at least the block doesn't have the same thing though and is instead much cleaner.



It's nice that the thermal pads are already attached for you - no cutting and measuring. As usual Swiftech have chrome plated the copper. The built in LEDs are powered by the fan header. This is nice and much cleaner than the designs that use LEDs connected to a molex. The LED's also light up the Swiftech logo on the bridge and a strip around the edge of the block:



Backside of the backplate has sections coated with plastic to avoid shorts even though clearance is fine:



The backplate is nice and thick and even has some ribbing around the IC's that it is designed to cool.

The top plate has a nice brushed aluminum look to it - and the removable section is where the gels for the LEDs go:



Sadly it's not possible to continue the brushed style onto the acetal/pom sections that are where the inlet/outlet ports are located. I do think it would be cleaner if the ports weren't raised up by design.



Instructions are clear enough and the pictures serve all languages. It would have been nice to have a bit of text though to just to clarify for noobs.



Fitted to block:







In action:





Now for Bitspower:







Nice to see so many spare parts though I've never needed a replacement o-ring on a GPU:



The block is simple and clean and the logos are more subtle. Overall quite handsome:



Note that there is an extra IC in the VRM section that BP cools that no one else does:







Sadly we didn't get the backplate to review - but it exists and it looks like this:



Fitted:





Watercool:













Super thin:






That's it for now. I'm hoping to wrap up the thermal this week assuming people stop asking for retests

[stren]

Alphacool backplate interfered with my RAM slots, result was that the card was only 95% of the way in the slot which meant the card is now dead. Proceeding to write up what data I have into full reviews but here's the short version:

Alphacool will have to wait for testing with the universal blocks after RMA. XSPC block had some trouble and I'm working with them to figure it out so that data is not considered good enough to publish yet. Swiftech also asked for a retest because of their VRM results, I'm not expecting anything different because the TIM spread was good on the backplate.

The pre-spiel:

- R9-290 tested at +100mV 1150MHz (203MHz overclock) on core, +100mV aux, 1375 (+125MHz) on Mem
- loaded with furmark, temps datalogged with aquasuite and hwinfo
- 4 hour warmup for TIM to burn in (MX2), then 20 minute averages for each flow point with 10 minute mini warm ups after changing each flow point
- 7 flow points representing minimum through to maximum flow rate from my MCP35x2
- 4770K @ 4.2GHz + Maximus VI Impact, Corsair AX1200/AX850 depending on the week, Corsair Dominator GT 2133 CL9-11-10 set to XMP, EX560 radiator with push pull yates, bunchs of Koolance QD4 and VL4N quick disconnects

Flow results:



Bitspower have stolen the "flow queen" award from Swiftech! The Koolance CPU-380 was included as a reference as it's a damn good cpu block and of "medium" restriction. Note how most GPU blocks are significantly more restrictive than it. The alphacool block is just terrible.

Thermal Results

This data is vs flow rate. It's hard to read so I tried to split this into block only and blocks with backplates to simplify things a bit. Block only:



Blocks with backplates:



I then took the average of all the flow points to make things easier to read. I also got rid of the other VRM data points as we're really only concerned with the main VRM temp sensor:



Analysis

Alphacool - $96 (Includes backplate)
- Thermal results TBD
- Awfully thick thermal pads (1.5mm, 2mm and 3mm used)
- Very high restriction
- Backplate can interfere with RAM slots on tight boards like the Maximus VI/VII Impact

Aquacomputer block - $123-142:
- Great core results
- Decent flow
- VRM results without the backplate were horrible - it felt like the screw was running out of thread though so I think this is due to poor contact

Aquacomputer backplate - $31-50:
- Passive backplate is awesome and a must have if you're going Aquacomputer because you do get great VRM cooling with it!
- Active backplate is probably not worth the money given that you're already better than every other backplate, but it does still add a little bit of extra performance (see this image for the results on this)

Bitspower - $140
- Very high flow
- Nearly 4C behind on core cooling - this is pretty disappointing
- VRM temps were "ok"

Bitspower backplate - $35
- Not reviewed, but does have thermal pads so it should cool the VRM even more

EK - $118-130
- Decent flow
- Decent core
- Decent VRM temps
- Good all rounder

EK Backplate - $29-37
- Better than a backplate with no thermal pads like XSPC, Watercool etc
- Not as good as the Aquacomputer backplate which is compatible with EK
- Available in multiple colors

Koolance - $130
- Good core temps
- Ok VRM temps
- No matching backplate
- Not very high flow

Swiftech - $180
- Backplate included
- High flow
- Best core temp results
- Let down by horrendous VRM temps despite the use of a backplate that tries to cool the VRMs

Watercool - $137-142
- Good core temps at low flow, but the block does not react as well to high flow as other blocks
- Good VRM temps (the best without a backplate)
- Medium flow

Watercool Backplate - $37
- Not reviewed, but no thermal pads so insulates the VRM rather than cools it

XSPC - $135
- Thermal Results TBD
- Not very high flow

XSPC Backplate - $40
- Not reviewed, but backplate has no thermal pads so insulates the VRM rather than cools it

Suggestions for next generation designs
- Thermal pads on backplates should be mandatory
- Core temps are pretty close these days with the exception of Bitspower
- VRM temps are still not close and manufactures can/should work on these
- Active backplates are the next battle that will separate the top tier blocks from the rest
- My gut feel is that the VRM chokes heat up the PCB considerably and that cooling these directly (or indirectly by thermal padding the backplate hugely) will help VRM temps significantly
- Block/backplate designs do seem to be getting more attractive with the exception of Alphacool


TLDR

- Money no object I would get the Aquacomputer block with active backplate. That combo is about as good as you can get, and it looks awesome.
- If money was a concern, I'd probably go with the EK as in copper/plexi form it's $118 and is a good all rounder
- The watercool block gets an honorable mention for VRM cooling with good low flow performance as does the Bitspower for high flow. The Koolance is also decent, but with no matching backplate is left out of the well dressed party.

[Jakusonfire]

Excellent work as usual Stren

Just a small suggestion. The block restriction chart is great and info that I find useful but the high range makes it hard to read detail in the low end. A zoomed in view around the 1 - 1.5GPM range would be useful for seeing the actual restriction levels in use rather than just the overall comparative view.

[Martinm210]

Simply amazing and exemplary work. Really sorry to hear you lost another card and the PSU, these VGA tests roundups are real hardcore hardware killers. I quoted your results over on OCN too just to get the word out on all your hard work and sacrifice!

I'm not even in the market for a new card or cooling, but find the results really interesting and satisfying, thanks!!

[stren]

Excellent work as usual Stren

Just a small suggestion. The block restriction chart is great and info that I find useful but the high range makes it hard to read detail in the low end. A zoomed in view around the 1 - 1.5GPM range would be useful for seeing the actual restriction levels in use rather than just the overall comparative view.

Yeah I've wondered about that - the signal levels are on the low end of my manometer so signal to noise is much lower, however as long as I take the data at higher levels then that should help keep accuracy up so that the best fit curve is still good. I know some people like to know the actual pressure levels at "normal" flow rates to know about how much "pump" they would need so I think that would be a good compromise.

Simply amazing and exemplary work. Really sorry to hear you lost another card and the PSU, these VGA tests roundups are real hardcore hardware killers. I quoted your results over on OCN too just to get the word out on all your hard work and sacrifice!

I'm not even in the market for a new card or cooling, but find the results really interesting and satisfying, thanks!!

Thanks Martin I'm sure you've had your own difficulties with OCN in the past so I'm not able to post my stuff there (though I haven't been banned quite yet). Yeah I always expect that manufacturers will have made everything the "same" by now but there are still differences and personalities to each. Good to see you around again though - did you finish your dunebuggy and house projects?

[Martinm210]

Yeah, they were a little overcontrolling a bit a while back but seem to be letting other members post roundup type reviews so I quoted it.

I'm still working on my buggy but after the second motor transplant, it is finally well into testing phase now:

Been taking friends and my kids out for spins:

A taste of central Oregon high desert with a friend:
http://youtu.be/vHCqB6nYki4

And Florence dunes with my daughter:
http://youtu.be/4JudG6UVEnc

I might get back to doing some PC stuff this fall but still having too much fun with the buggy stuff. Still got some minor stuff to do, but it's fairly solid now and getting close to tear down/paint phase.

[m_jones_]

Thank you as always for your great results, sad to hear about your hardware failures.

[stren]

Yeah, they were a little overcontrolling a bit a while back but seem to be letting other members post roundup type reviews so I quoted it.

I'm still working on my buggy but after the second motor transplant, it is finally well into testing phase now:

Been taking friends and my kids out for spins:

A taste of central Oregon high desert with a friend:
http://youtu.be/vHCqB6nYki4

And Florence dunes with my daughter:
http://youtu.be/4JudG6UVEnc

I might get back to doing some PC stuff this fall but still having too much fun with the buggy stuff. Still got some minor stuff to do, but it's fairly solid now and getting close to tear down/paint phase.

Nice work that looks pretty darn cool!

Thank you as always for your great results, sad to hear about your hardware failures.

Yeah only one motherboard, one PSU and two GPUs lulz. And the manufacturer's complain that it takes me so long to complete this stuff haha. The motherboard is the only one I went out of pocket on so far. Hopefully Sapphire will RMA the GPU again though it's hardly their fault.

Excellent work as usual Stren

Just a small suggestion. The block restriction chart is great and info that I find useful but the high range makes it hard to read detail in the low end. A zoomed in view around the 1 - 1.5GPM range would be useful for seeing the actual restriction levels in use rather than just the overall comparative view.

So I did some trimming as you suggested - two versions:

[Church]

Hmm, imho indeed no need to go higher then 1.5gpm. But it may be worth to start from 0.2-0.25gpm, as flow per block can get rather low for SLI/CF paralel loop setups.

[Jakusonfire]

That's great Stren. Thank you for that. Aquacomp have made major improvements to their flow restriction this round but its still interesting that some blocks are twice the restriction of others.
XSPC I think was one of the least restrictive last round but their 290X is reasonably high.

[stren]

Hmm, imho indeed no need to go higher then 1.5gpm. But it may be worth to start from 0.2-0.25gpm, as flow per block can get rather low for SLI/CF paralel loop setups.

You mean for restriction testing or temp testing? I'm limited on the low end of flow for temps both by how low I can turn the pump down to, and how low my flow meter can read I don't know I would just mentally continue the curves of best fit back Given the slope of the watercool block it would probably overtake the others at very low flows.

That's great Stren. Thank you for that. Aquacomp have made major improvements to their flow restriction this round but its still interesting that some blocks are twice the restriction of others.
XSPC I think was one of the least restrictive last round but their 290X is reasonably high.

Yeah I went and dugout the old data from titan and adjusted the scale to match:



XSPC - 5PSI (titan) vs 6PSI (290) @ 2GPM
AquaComputer - 7.9PSI (titan) vs 3.3PSI (290) @ 2GPM

So not too different on XSPC - XSPC have confirmed they used the same cooling engine design as the titan block so you'd expect restriction to be similar. It's possible I used different ports when testing flow on Titan as the XSPC has 7 port options, 2 of which remove the 90 degree bends that I used on the 290 testing.

Wrote up the full thing on the website: http://www.xtremerigs.net/2014/08/11...block-roundup/

Doing a bit of CPU block testing now, and then once the card gets back from RMA I'll be doing some universal GPU blocks.

[Aquatuning]

It's a shame about the Alphacool block causing clearance issues but we have been told that they are revising the back plate and that they will send you one to complete the tests.

Regarding your point about the thickness of the pads being 3mm to make sure the soldered tips do not touch the back plate its actually a growing problem we are seeing this a lot more with recent generations of GPU's. It's also the most user friendly method of making sure there is not an issue unless you use a method like this from Swiftech that is not exactly friendly for the customer.

[stren]

It's a shame about the Alphacool block causing clearance issues but we have been told that they are revising the back plate and that they will send you one to complete the tests.

Regarding your point about the thickness of the pads being 3mm to make sure the soldered tips do not touch the back plate its actually a growing problem we are seeing this a lot more with recent generations of GPU's. It's also the most user friendly method of making sure there is not an issue unless you use a method like this from Swiftech that is not exactly friendly for the customer.

While I agree that Swiftech's method is undesirable - Aquacomputer use 0.5mm thermal pads in specific areas without issue. Therefore if Alphacool wanted to be conservative they could use 1mm or 1.5mm. This however requires the machining of the backplate with some accuracy. I do not believe that the thermal pads *need* to be 3mm to avoid this nor do I believe that Aquacomputer, EK or Bitspower are unsafe. Engineering a better solution is better than brute forcing one. There are specific areas where through hole components are used, and these can be avoided by design

[SuperMiguel]

is the aquacomputer one compatible with the MSI 290x gaming edition the twin frozr version?

[stren]

is the aquacomputer one compatible with the MSI 290x gaming edition the twin frozr version?

I don't believe so but I don't have one to try. I believe the EK is and *maybe* some of the others that got updated later.

[stren]

So just to correct some misinformation that I was told - I may have mentioned being told that jayztwocents had problems with the Alphacool backplate also - he confirmed to AT that this was not the case, so I would like to correct that. They have also altered the design of the backplate and are sending me another!

I have the GPU back from RMA again so hopefully we will have new results soon!

[Martinm210]

Your a trooper continuing with this. No way would I risk a second loss, let alone a third. Then again, I'm too cheap to even buy the first!!lol

[Johnny87au]

Your a trooper continuing with this. No way would I risk a second loss, let alone a third. Then again, I'm too cheap to even buy the first!!lol

awesome testing bro, honestly i don't like the look of the AC block, results are great but i'd have to go on cost/aesthetics/performance..