Swiftech Releases Komodo HD5800 F/C Waterblock For ATI

[johnnymossville]

dejanh, Thanks for the tips. I'm going to be very methodical when doing this and will post my findings here.

[dejanh]

Received the standoffs. One standoff had a tapping fault and broke one of my fasteners and another standoff came loose when I was trying to remove the block after initial testing and knocked off a corner of one of the parts but no contacts were broken so the card still works...sucks though The rest seem fine.

Currently trying 0.5mm reduction on main VRM area and 0.25mm reduction on the single VRM at the front of the card. The temperatures are more-less under control, however by no means great. I am not seeing consistent improvement as the contact is not consistent. At 1.262V the VRM temperatures range from ~70C (decent contact) to ~100C (poor contact) So far I am seeing two issues...

(1) The fasteners cannot be fully screwed into the standoffs as they are too long for the standoff thread depth now that I am using shorter standoffs around VRM area(s).

(2) The other standoffs, particularly the one near the rear main VRM area do not allow the PCB to be compressed much further to the block resulting in uneven block contact with the VRMs.

[Prometheus2010]

@Dejanh, Thanks for the update! It looks like my last option will be replacing the thermal pads on these VRMs. I just bought EK thermal pads .5mm and 1mm thickness, not sure what's the original's, but hopefully this will bring the VRM temp back to normal. If not, then I'll have a very expensive paperweight

[ibuycheap]

@Dejanh, Thanks for the update! It looks like my last option will be replacing the thermal pads on these VRMs. I just bought EK thermal pads .5mm and 1mm thickness, not sure what's the original's, but hopefully this will bring the VRM temp back to normal. If not, then I'll have a very expensive paperweight

Lmao I just did the same thing. I'm not using the komodo but the swiftech hd5850 heatsink and it has the same issue. The standoffs are too high. 2 of the ram chips make little contact and the vrms barely any contact.

[dejanh]

Testing on hold again...turns out the card I just received is defective and needs to go back. It constantly keeps locking up under 3D load. Got to wait until Monday to see about getting it replaced.

This is so damn annoying now...I am getting quite irritated with all of the assembling and dissasembling. I guess another round of taking down the loop is coming up

[gabe]

@Dejanh, Thanks for the update! It looks like my last option will be replacing the thermal pads on these VRMs. I just bought EK thermal pads .5mm and 1mm thickness, not sure what's the original's, but hopefully this will bring the VRM temp back to normal. If not, then I'll have a very expensive paperweight

NO you won't.. you can always call us for support you know?

[gergregg]

NO you won't.. you can always call us for support you know?

That right there is great customer service!

[dejanh]

Look guys, let me make this very clear...

Pads work just fine on this block. The VRM temperatures with pads are far from optimal, but they are under control and can hold a very decent OC just fine. Having said that, on to my results...

Mounting the block with paste at VRM areas instead of pads does significantly reduce the VRM temperatures as stated by Gabe, up to 20C in fact. The standoff reduction that is needed around main VRM area is 0.5mm, and around the single VRM at the front of the card minimum 0.25mm. It also requires full compression of the PCB to the standoffs. However, this is not without issues.

I reported earlier that my card appears to be locking up and I thought that the card was faulty. However, it turns out that was not the case and in fact it was the reduced standoffs that were causing the problem, specifically, it was the board warping that happens as a result of the use of reduced standoffs that was causing random lockups on 1 out of 2 cards and was resulting in no video post or loss of video in some cases.

Compressing the PCB all the way down to the standoffs is a must if you want to make proper contact with the block when using paste on VRMs and this causes significant warping of the board both lengthwise and widthwise. In fact, widthwise the PCB starts to take on a bow-shape. As a result, only the outer VRMs make good contact while the ones in the center (two large and one small) make poor contact again. There is no warping around the single VRM at the front of the card. The warping is the result of two things. First, the rest of pads on the block used for memory, that concave cut near the VRMs, etc. get in the way and need to be removed and replaced with paste as well. Second, the rest of the standoffs are too tall resulting in warping and need to be equally reduced by 0.5mm as removing the rest of the pads also means that the rest of the board components will not make contact any more with the block.

So, what does this all mean? Well, it really means only one thing. If you want to use paste two things need to happen...

1. All pads need to be removed on the whole block and must be replaced with thermal grease (paste).
2. Gabe supplies all new set of standoffs for the whole block reduced by 0.5mm and supplies a set of new mounting screws to go with the reduced standoffs OR you reduce your own standoffs by 0.5mm which I do not really recommend as inaccuracies can cause more problems than good.

As it is right now, this block is not reliably and consistently suitable for use with HD5850 cards when using paste method and this method should therefore not be used.

If Gabe supplies me with standoffs reduced as per my findings above I will attempt to redo the tests again and post back the results. For now I am back on pads.

[johnnymossville]

I just got my rig together tonight. I have some cleaning up to do inside the case and some tweaking to the bios, etc,... but I thought I'd run a furmark test to see how the komodo with my 5870 is doing. This is my first attempt. It's still running, over 10 minutes now, and the vrm's haven't gone over 59c.

I'm very happy with it for now, and know it's still got more left in it after some minor tweaking.

[Philwong]

I just got my rig together tonight. I have some cleaning up to do inside the case and some tweaking to the bios, etc,... but I thought I'd run a furmark test to see how the komodo with my 5870 is doing. This is my first attempt. It's still running, over 10 minutes now, and the vrm's haven't gone over 59c.

I'm very happy with it for now, and know it's still got more left in it after some minor tweaking.

Are you using pad or grease?

Phil

[johnnymossville]

Are you using pad or grease?

Phil

I'm using thermal grease. I also ran into the shorting/no vga output problem on the back of the card. I did the back-off solution Gabe mentioned and that worked.

[dejanh]

I just got my rig together tonight. I have some cleaning up to do inside the case and some tweaking to the bios, etc,... but I thought I'd run a furmark test to see how the komodo with my 5870 is doing. This is my first attempt. It's still running, over 10 minutes now, and the vrm's haven't gone over 59c.

I'm very happy with it for now, and know it's still got more left in it after some minor tweaking.

Are you using pad or grease?

Phil

Use FurMark 1.8, Xtreme Burning Mode, set the resolution to at least 1680x1050, and crank that voltage up to at least 1.25V with a decent overclock. I am curious to see what happens then. Those VRM temps you are posting are not that impressive considering you are running stock voltages.

Having said that, I have been testing yet another proposed solution from Gabe and I am happy to report that the results are looking good. I am going to let Gabe know today what the results are and then I will post them here together with the solution. While my VRM temperatures are still high they are under control and, I can confirm that paste works for me now as well since I have dropped an average of 20C from the peak VRM temperatures with the new test setup.

I will post the details and requirements of the new setup, i.e., what you need to make it work later today.

[johnnymossville]

Will do dejanh, and good to hear you are getting those vrm temps under control. you had some scary numbers going there.

I'm going to have to get a different bios from Powercolor, all they let me do with the installed bios is 900/1300. More than enough for me, but I do want to push it and refine it a bit more before I settle into an economical 24/7 setup.

I haven't played with overclocking other than to set it to 4.2ghz and set the memory to DDR2000/CAS8. It's been running since I first powered it yesterday.

I'm a 3D designer/animator, so realtime graphics performance takes a second seat to Render speed for me. I'm no overclocker that's for sure. haha

having fun though!

[dejanh]

Will do dejanh, and good to hear you are getting those vrm temps under control. you had some scary numbers going there.

I'm going to have to get a different bios from Powercolor, all they let me do with the installed bios is 900/1300. More than enough for me, but I do want to push it and refine it a bit more before I settle into an economical 24/7 setup.

I haven't played with overclocking other than to set it to 4.2ghz and set the memory to DDR2000/CAS8. It's been running since I first powered it yesterday.

I'm a 3D designer/animator, so realtime graphics performance takes a second seat to Render speed for me. I'm no overclocker that's for sure. haha

having fun though!

Fur sure I have had my share of stress though as well. Testing this when you are paying out of your own pocket is no joke. Parts die and nobody finances my replacements.

Having said that, I would be really curious to see your results with FurMark 1.8 using Xtreme Burning Mode, if you can 1920x1200 resolution, ~1.3V on the card with 1000MHz core and 1250MHz-1300MHz on the memory.

[johnnymossville]

...I would be really curious to see your results with FurMark 1.8 using Xtreme Burning Mode, if you can 1920x1200 resolution, ~1.3V on the card with 1000MHz core and 1250MHz-1300MHz on the memory.

I might work my way up to that as I gain some courage. Raw performance is cool and Dirt2 totally kicks arse on this machine, but I think the nicest part of water cooling the gpu is how quiet it is compared to the stock cooler. That thing sounds like a jet taking off!

And yes, failure can be expensive if things go bad. I haven't built up a pc in a few years before this one, and I definitely had a few moments of stress when things weren't going right along the way. All kinds of thoughts run through your head when things go bad!!! LOL

[Xtrykr]

Look guys, let me make this very clear...

Pads work just fine on this block. The VRM temperatures with pads are far from optimal, but they are under control and can hold a very decent OC just fine. Having said that, on to my results...

Mounting the block with paste at VRM areas instead of pads does significantly reduce the VRM temperatures as stated by Gabe, up to 20C in fact. The standoff reduction that is needed around main VRM area is 0.5mm, and around the single VRM at the front of the card minimum 0.25mm. It also requires full compression of the PCB to the standoffs. However, this is not without issues.

I reported earlier that my card appears to be locking up and I thought that the card was faulty. However, it turns out that was not the case and in fact it was the reduced standoffs that were causing the problem, specifically, it was the board warping that happens as a result of the use of reduced standoffs that was causing random lockups on 1 out of 2 cards and was resulting in no video post or loss of video in some cases.

Compressing the PCB all the way down to the standoffs is a must if you want to make proper contact with the block when using paste on VRMs and this causes significant warping of the board both lengthwise and widthwise. In fact, widthwise the PCB starts to take on a bow-shape. As a result, only the outer VRMs make good contact while the ones in the center (two large and one small) make poor contact again. There is no warping around the single VRM at the front of the card. The warping is the result of two things. First, the rest of pads on the block used for memory, that concave cut near the VRMs, etc. get in the way and need to be removed and replaced with paste as well. Second, the rest of the standoffs are too tall resulting in warping and need to be equally reduced by 0.5mm as removing the rest of the pads also means that the rest of the board components will not make contact any more with the block.

So, what does this all mean? Well, it really means only one thing. If you want to use paste two things need to happen...

1. All pads need to be removed on the whole block and must be replaced with thermal grease (paste).
2. Gabe supplies all new set of standoffs for the whole block reduced by 0.5mm and supplies a set of new mounting screws to go with the reduced standoffs OR you reduce your own standoffs by 0.5mm which I do not really recommend as inaccuracies can cause more problems than good.

As it is right now, this block is not reliably and consistently suitable for use with HD5850 cards when using paste method and this method should therefore not be used.

If Gabe supplies me with standoffs reduced as per my findings above I will attempt to redo the tests again and post back the results. For now I am back on pads.

Just out of curiousity, is it possible to sand down the standoffs or do they crack/break?

[gabe]

Having said that, I have been testing yet another proposed solution from Gabe and I am happy to report that the results are looking good. I am going to let Gabe know today what the results are and then I will post them here together with the solution. While my VRM temperatures are still high they are under control and, I can confirm that paste works for me now as well since I have dropped an average of 20C from the peak VRM temperatures with the new test setup.

good to hear from you, and that the new screws worked out well. looking forward to see your new test results.

[dejanh]

good to hear from you, and that the new screws worked out well. looking forward to see your new test results.

Sorry, I've been completly busy and unable to prepare results. I did test though already just quickly for myself as soon as I received the parts from you and everything worked out. I would suggest that you include the revised mounting with all the blocks. I can tell you that tightening the screws 1/2-1 turn more can result in 10C-15C reduction in temperatures. I originally thought that it was not working again on one card but then I started tightening the blocks further and achieved very decent results.

The block can actually hold 1.35V now and >1000MHz overclock on both cards in CrossFireX.

[hoodlikegaza]

so what's the verdict? Is the komodo going to start shipping with these standoffs and screws?

[gabe]

so what's the verdict? Is the komodo going to start shipping with these standoffs and screws?

There is no change in the standoffs. Just the screws. We will send replacement screws upon request for any product purhcased in the channels that are using the old screws, and roll-out the new screws for the new shipments.

[dejanh]

I did not include a screenshot here because FurMark crashed just as I was about to take the screen but I can tell you that at 1.35V, 1030MHz/1250MHz with two HD5850s hooked up in CrossFireX with parallel water flow the peak temperatures on the VRM are 98C on card #1 (top slot), and 90C on card #2. The temperatures range a lot, with the hottest temperatures being observed on the middle VRM of the 3 since this is the hardest one to make contact with as there are no fasteners around the middle of the card width-wise. The two outer VRMs are both lower by 10C - 15C. Therefore, range is from 70C to 98C on all of the VRM on both cards. Tightening the screws futher does help lower the temperatures a bit still, but this is okay since the temps are completely under control now. There are no more runaway thermals.

All of the tests were done using FurMark 1.8 Multi-GPU, Xtreme Burning Mode, at 1920x1200 with no AA, Post Processing enabled. The load was consistent at 97% on each card, peak power draw was ~90 amps/120W, average was ~85 amps/115W. This power load will never be observed under normal operating conditions and under normal conditions the VRM temperatures never exceed 55C-60C from what I noticed.

[johnnymossville]

Dejanh: What kind of frame rates were you getting with those settings and with crossfire?

John

[dejanh]

Dejanh: What kind of frame rates were you getting with those settings and with crossfire?

John

Pretty low, I think about 40FPS.

[Philwong]

All of the tests were done using FurMark 1.8 Multi-GPU, Xtreme Burning Mode, at 1920x1200 with no AA, Post Processing enabled. The load was consistent at 97% on each card, peak power draw was ~90 amps/120W, average was ~85 amps/115W. This power load will never be observed under normal operating conditions and under normal conditions the VRM temperatures never exceed 55C-60C from what I noticed.

So far, only Crysis is capable of heating up my 5870's VRM to 90c+ at 1,920 x 1,200 reso. For most games such as Dirt 2, COD:MW2 and CNC4, the temp is around 65c peak. Nevertheless, good job on all the beta testing.

Phil

[Philwong]

There is no change in the standoffs. Just the screws. We will send replacement screws upon request for any product purhcased in the channels that are using the old screws, and roll-out the new screws for the new shipments.

Gabe, I don't quite understand how the screws can remedy the contact issue instead of the standoffs. Could you please elaborate?

By the way, Michelle will be sending me some missing screws and spacers for the Apogee GT/AM2 plate I ordered from PPCS. Can I request the revised screws for my Komodo 5800 to be shipped over as well? Thanks.

Phil