Koolance 8800 Block: Pics and Evaluation

Hey,

This is an evaluation of Koolance's new 8800GTS block. You'll find pictures below that show the inner channeling of the block, along with an evaluation of my experience with this cooler. I'll be posting performance data just as soon as I can drain the cooling loop in my rig and get this installed. I'm also going to throw in my opinion of the block itself on the basis of design and build quality. As a side note, while this is the GTS block, I believe it's safe to assume the channel design of the GTX block is the same.

On to the pics...

Out of the box:




Interface surface - look closely and you'll observe the more than noticeable markings where the mill bit passed by. It would have been nice if Koolance had taken the time to properly flatten the surface, either with machine or hand lapping, before the gold plating went on. You wouldn't see this on an EK, DangerDen or Swiftech product.




Screws out:




Top side of cover plates:




Bottom side of cover plates:




Inside the block. Decent channeling, though maybe not as deep as the competition. I did not measure and compare it to something else, so keep in mind it's just my observation. The square pin grid looks pretty good. It appears to offer more cooling surface area than many other full cover blocks, save for the EK. I've not seen the inside of DD's 8800 blocks, but I know this looks better than my Koosah. The pin grid should create a lot of turbulence too, and I like the fact it's roughly the size of the 8800 GPU, not just a small focused area in the center. The seal is a clear silicone type material, and is held lightly in place by a thin grease. Seems adequate.






I'm replacing the stock 3/8" Koolance fittings with Perfect Seal 1/2" fittings from DD. This shows how far the fitting protrudes into the channel, so I'll be spacing it out (see below):




Close up of the pin array:




Perfect Seal fitting with bronze washer and thicker o-ring. It fits tight to the threads, so I'm hoping it won't leak.




Fittings installed:
















I'll be mounting it to the card next. Following that, I'm going to drain my cooling loop, install the card, and plumb this in. Of course it's going to take a couple days to bleed the loop, so I don't expect to have any performance figures posted until that time. I will keep the thread up to date with pictures as things progress.


Tom

Wanted to note a couple things that I didn't in the original post.

1. I opted not to reuse the top Koolance cover plate. The new thicker o-ring won't clear the holes in the plate - I could drill them, but I'm just as happy to leave it off. The holes in the main block are deep enough that the screws don't bottom out before sealing it.

2. I do have some concern about the Al top cover, as the rest of my loop has copper components. I'm running MCT-5, which should inhibit galvanic corrosion between the metals, but it would have been best with a copper, acetal or even an acrylic top.


On to the installation...

looks like a great block, i will be interested to hear how it performs!

Installation pics...

eVGA 8800GTS, stock cooler removed:




Cleaned:




Bare G80 GPU:




Koolance TIM pads for Vreg components:




Bare GDDR3 chips:






Prepped with AS5 on the GPU, generic thermal paste on RAM and PCI-e controller:




Installed, back side:




Front side:




I checked contact points, and I could see light passing between the block and RAM chips. The block doesn't come nearly close enough to the RAM to make good contact with thermal paste applied as thin as one should. I removed the block, cleaned the miniscule bits of paste that actually did contact the block, and cleaned the RAM chips all over again with alcohol.

Koolance supplies 8 pieces of pre-cut thermal interface material, 4 apply to the Vreg chips (see above), 3 are used for the RAM, and 1 for the PCI-e controller. I applied the 3 RAM pieces, but left the PCI-e controller with thermal paste as it does make proper contact with the block. After re-attaching the block to the card, could see the TIM pads had closed up the space between the block and the RAM. I'm really wondering if it would be better off not being there at all...but I figure it can't really hurt either. I expect the competing full cover blocks will make full contact with the RAM. Whether that's a benefit or not, I don't know, but I can state that it makes no difference with the RAM on my 7800GT - the RAM tops out at the exact same clock whether they're free air or watercooled with the DD Koosah.

The assembly will be installed and plumbed into my rig tomorrow. I'll report back tomorrow night or Monday with some results...


Tom

Wow, that chip is HUGE.

Looks good though with the waterblock, makes the card look much more professional IMO

Click to expand...

Is it me or is the RAM of the bottom right out of Socket?

Xerodeluxe said:

Is it me or is the RAM of the bottom right out of Socket?

Click to expand...

Are you referring to the two empty spots to the lower right of the GPU? If so, it's because they are 64MB chips, and this is a GTS...10x64MB=640MB. If there were a full 12, we'd have 768MB for the GTX.

Or are you seeing something else that I should be worried about?

i await your results, i am contemplating the gtx version for my new build

hemi said:

Are you referring to the two empty spots to the lower right of the GPU? If so, it's because they are 64MB chips, and this is a GTS...10x64MB=640MB. If there were a full 12, we'd have 768MB for the GTX.

Click to expand...

Yup, I think thats what he is referring to.

Xerodeluxe said:

Is it me or is the RAM of the bottom right out of Socket?

Click to expand...

Hahaha. Yeah, it's not supposed to go there . I thought that too when I first saw a board shot.

I'm wondering how this block will perform, since there seems to be a LARGE pin array over the entire IHS of the card, yet there's VERY little surface area DIRECTLY over the core.

The GTS gets about 46 Load with the MCW60 + adapter plate.

Arcygenical said:

Hahaha. Yeah, it's not supposed to go there . I thought that too when I first saw a board shot.

I'm wondering how this block will perform, since there seems to be a LARGE pin array over the entire IHS of the card, yet there's VERY little surface area DIRECTLY over the core.

The GTS gets about 46 Load with the MCW60 + adapter plate.

Click to expand...

Cool, thanks for the info - is that load temp at stock, or your clock of 675? 46C sounds pretty good...not sure if my loop has the cooling capacity of yours. Obviously no real way to get a direct comparison because of all the variables, but I'm curious to see where this ends up.

I'll do my best to get you guys some preliminary numbers tomorrow evening while the loop is bleeding.


Tom

Interesting to see what the results are goin to be. Nice block, but I really don't think the gold plating does anything in the way to help performance, I mean if u want some kind of material that really helps, than they should be using synthetic sheets of diamond, which is really not that expensive. FYI, diamonds thermal interia or conductivity is 44,000 times that of copper! Its real hard to get those carbon atoms in the diamond tetrahedron crystal lattice to seperate.

Caliche said:

Interesting to see what the results are goin to be. Nice block, but I really don't think the gold plating does anything in the way to help performance, I mean if u want some kind of material that really helps, than they should be using synthetic sheets of diamond, which is really not that expensive. FYI, diamonds thermal interia or conductivity is 44,000 times that of copper! Its real hard to get those carbon atoms in the diamond tetrahedron crystal lattice to seperate.

Click to expand...

Yeah, gold actually inhibits heat absorption.

its so thin that i doubt it has any measurable affect. I think it was a smart move considering thier rads are aluminum, as is the top plate of this block.

spint3 said:

its so thin that i doubt it has any measurable affect. I think it was a smart move considering thier rads are aluminum, as is the top plate of this block.

Click to expand...

Then why didn't they coat the aluminum block too?

Minor setback. I got it installed and filled the loop, but the top barb on the block is seeping a little around the smaller o-ring when the pump is running and the loop is under pressure. I am going to attempt to source a couple replacement fittings today, but if not, I'll have to wait a couple days for 1/2" Koolance fittings to arrive in a couple days. The block itself and the lower barb sealed well.

So close...it's not a big deal, but I'm still salty that I may not be able to get it running today. Oh well.


Tom

nice enough block waiting to hear what kind of temps you can get with it?????
personally have oced gtx with a EK waterblock and i'm having pretty dang good temps with it, lets see can this koolange block beat my temps: idle 39C load 42c.

Arcygenical said:

Then why didn't they coat the aluminum block too?

Click to expand...

beats me, but they must have some reason for only coating thier copper products.

spint3 said:

beats me, but they must have some reason for only coating thier copper products.

Click to expand...

I wonder if aluminum can be effectively electroplated with Gold.

I can see where they're coming from though, copper DOES oxidize in contact with ambient air.

Solarfall said:

nice enough block waiting to hear what kind of temps you can get with it?????
personally have oced gtx with a EK waterblock and i'm having pretty dang good temps with it, lets see can this koolange block beat my temps: idle 39C load 42c.

Click to expand...

Me too, can't wait to get it together. Those numbers are stout - stock clocks? I really wanted the EK GTS acetal top block, and I might still get one. Hopefully they are back in stock soon.

Arcygenical said:

I wonder if aluminum can be effectively electroplated with Gold.

I can see where they're coming from though, copper DOES oxidize in contact with ambient air.

Click to expand...

So does aluminum - almost instantly. The top plate for the block seems anodized, but I don't know if it's [H]ard anodized (sorry, couldn't resist...but that was supposed to be serious).




No luck finding fittings today. I'm going to have to order some. I'd really like to score a couple EK barbs with the spacers that have the 0-ring receiver groove, but I can't find them separate. The Koolance pieces will have to do. EDIT: Woohoo! Found the G 1/4" EK spacers at Petra's.

Here are a couple more pics. It's about 1/4" longer than the 7800GT it replaces. FWIW, I also must mention the 8800GTS with the Koolance block is noticeably heavier than the 7800GT with the DD Koosah

I gotta say, the gold plating against the black PCB's of the card and the mobo is pure sex - hopefully it lives up to it in the performance department.







Since I was bored and sobbing today, I made a couple widescreen wallpapers of the G80 chip from pics I took (I plan on using them for my desktop). Clicking the thumbs below will open 1920x1200 versions, or you can of course do the "right-click, save link as/save target as" action instead if you so choose.








Be back in a few days with an update...I'm out..

So does aluminum - almost instantly. The top plate for the block seems anodized, but I don't know if it's [H]ard anodized (sorry, couldn't resist...but that was supposed to be serious).

Click to expand...

I'd HOPE it's hard anodized...

Update...

I was able to get it set up leak-free with a set of Danger Den barbs that didn't have as much thread reach as the "Perfect Seal" barbs. I am typing this from the machine as it finishes bleeding air, but I have some preliminary temps to share...nothing earth shattering, or not as good as I expected, but it's still early.

Stock clock is 513/1584
Idle: 49C(!), Peak: 61C(!)

My Conroe's temps are up 5C at idle. This G80 dumps some heat for sure.

The AS5 still needs to cycle in, and the air needs to finish bleeding out of the loop...but I doubt it's going to make more than 4-5C difference, if that.

The guys who have better numbers with the Swiftech and EK blocks, can you post your configs for comparison? (GPU/RAM clocks, TIM, heat exchanger (rad), fan speed & approx CFM, pump, tubing size)

My loop:

Coolingworks Coolrad 22T
2x Scythe S-Flex 120mm fans @ 7V (800RPM = 38CFM approx.)
Laing D5, speed 4
1/2" tubing

I could crank the fan speed up, but I'm into quiet computing as much as I'm a power user, so I'll leave it as is. I'll post back when things settle in and I crank up the clocks. If this doesn't pan out (and it's not looking so great at the moment), I'm going to order an EK GTS block.


Tom

whsts you cpu temps?

8800GTS @ 675/2100 w MCW60 46 Load

Conroe E6300 @ 3.45 (for now)
BIP Xflow with 4 7v Yate Loons Push/Pull
DD TDX waterblock
MCP655

Chipset is Watercooled also

hemi said:

Me too, can't wait to get it together. Those numbers are stout - stock clocks? .

Click to expand...

No.... slightly overclocked. now running: core 650mhz /mem 1053 mhz. havent had a lot of time fidle around the settings yet, but i bet there's still some head room left in my XFX.
BTW it's not the factory oc:ed xxx model.

Arcygenical said:

8800GTS @ 675/2100 w MCW60 46 Load

Conroe E6300 @ 3.45 (for now)
BIP Xflow with 4 7v Yate Loons Push/Pull
DD TDX waterblock
MCP655

Chipset is Watercooled also

Click to expand...

How much voltage do you have going to the CPU? What's your ambient/intake temp?

I'll post some more numbers tonight, but I took it to 650/2000 last night and the temps didn't go much higher than @ stock. Still not good, IMHO.

@Chilly: Idle was once ~40C (sometimes lower depending on ambient), now it's 45C with the 8800. Peak is still low to mid 50's, but I haven't checked the CoreTemp log to confirm that. It's an E6600 at 3.5GHz with 1.52v (does 3.6 @ 1.56+, but isn't 24/7 stable).

@Solarfall: They are some great numbers. Can you share the details of your cooling setup and your other components? (ambient temps help too)

There's no good way to determine if I have another problem in my loop or if it's the Koolance block itself. I can say I was running everything else the same and my 7800GT with the Koosah idled mid 30's and peaked about 41-42. Granted, it isn't nearly the power hog the 8800's are, so maybe my heat exchange is no longer keeping up with the load.

BTW, my 8800 has the squealing problem I've read about. It's definitely coming from the power regulation on the board. Is that normal and something you guys are experiencing?


Tom

hemi said:

I'll post some more numbers tonight, but I took it to 650/2000 last night and the temps didn't go much higher than @ stock. Still not good, IMHO.

Click to expand...

That actually sounds good to me, my stock clocked air cooled 8800GTS idles at 45C, load up to 70C sometimes, mostly mid/high 60s. If you can get those temps OC'd with water that doesn't sound too bad.

theelectic said:

That actually sounds good to me, my stock clocked air cooled 8800GTS idles at 45C, load up to 70C sometimes, mostly mid/high 60s. If you can get those temps OC'd with water that doesn't sound too bad.

Click to expand...

Yeah, I guess it's keeping the peak a little lower (but idle is higher), but it's not even in the same league as Arcygenical or Solarfall. I suppose I'm also benefitting from the fact it's near silent (other than the squeal), but I'd bet Arcygenical's quad Yates @ 7v aren't much louder and his cooling performance is much better.

We'll see where it goes...


Tom

hemi said:

How much voltage do you have going to the CPU? What's your ambient/intake temp?

Vcore (everyday usage 3450mhz) = 1.36275v according to my bios
Vcore for the 3500mhz Stable run was 1.304v.

it crashed after 9 hrs Orthos, so I bumped up the Vcore and lowered the OC a bit.


Tom

Click to expand...

hemi said:

Yeah, I guess it's keeping the peak a little lower (but idle is higher), but it's not even in the same league as Arcygenical or Solarfall. I suppose I'm also benefitting from the fact it's near silent (other than the squeal), but I'd bet Arcygenical's quad Yates @ 7v aren't much louder and his cooling performance is much better.
Tom

Click to expand...

I've measured my Quad yates at 25dba during a LOAD operation, and 21dba during idle. I have to manually adjust the fans though. If I leave them on low for my load runs I get temperatures in the low 60s on the CPU and 52c on the GPU.

I'd trade my block for a full coverage ANY day... Too much messing around with all the Ramsinks and stuff... Infact, I plan on buying a GTS waterblock with the cash I'm making from selling my Powermac.

I just wanted to say thanks for this helpful post. I've been looking at all the block options for cooling mine. Hopefully this block turns out half way decent. So far your temps are better then what I'm getting on air. Also, thanks for the new wallpapers! Keep up the good work.

Arcygenical said:

I've measured my Quad yates at 25dba during a LOAD operation, and 21dba during idle. I have to manually adjust the fans though. If I leave them on low for my load runs I get temperatures in the low 60s on the CPU and 52c on the GPU.

I'd trade my block for a full coverage ANY day... Too much messing around with all the Ramsinks and stuff... Infact, I plan on buying a GTS waterblock with the cash I'm making from selling my Powermac.

Click to expand...

I hear you - no way I was going to mess with cooling all those individual components. I'm damn near salivating over the EK GTS block. We'll see where things go...

Sta!nless said:

I just wanted to say thanks for this helpful post. I've been looking at all the block options for cooling mine. Hopefully this block turns out half way decent. So far your temps are better then what I'm getting on air. Also, thanks for the new wallpapers! Keep up the good work.

Click to expand...

No problem, and thanks!

Another update...

I bumped the D5 (aka MCP655) up to speed 5 from 4, just to see what flow would do. Doesn't seem like anything.

Since I hadn't actually fully stressed the card after bumping up the clocks, I decided to run rthdribl again as I did when stock. The card is hovering around a peak of 66-67C at 650/2000, running rthdribl in a 1680x1050 window (see screenshot below).

Using an Omega digital temp probe with a thermocouple, I measured the intake, core and exhaust temp of the heat exchange. The intake temp with the thermocouple at approximately 1" from the radiator core was 24C. The core itself was 29C, and the exhaust temp was a steady 32.4C.




Click for fullsize...


I'll continue testing and try a few more things. Of note is that I measured the side of the block next to the fittings, and the temp is ~34C. It seems the liquid is pulling heat away from the block, but I don't know how good a job the block is doing pulling heat from the GPU. After I complete testing, I'm considering pulling the block off the card and checking for proper contact. The AS5 still needs some time, but the surface irregularities I mentioned up in the original post may be at play.


Tom

Last update for the day...

I've cranked the radiator fans up to 12V (~63CFM each), and lowered my C2D to 3.3GHz @ 1.37v. The GPU now idles at 46C. I ran rthdribl for 30 minutes, again at 1680x1050, and it fluctuated between 59 and 60C from ~4 minutes in until the end. The CPU, which was sitting at 54/51 in the screenshot above, now tops out at 41/40 while running rthdribl and idle temp has dropped from 46/44 to 37/35. That said, the point was to take the CPU's heat load out of the equation, but keep it somewhat realistic. Obviously I could have dropped the CPU to stock and undervolted it, but there's no way I'd run it long term that way with this rig...

Anyway, it seems this block just isn't working out. Even running the fans full tilt and lowering the heat dump of the CPU to net a 7C drop in GPU load temp isn't adequate in my mind when it's still topping out at 60C. I'm not sure what else to try, so I'm open to suggestions. At this point in the game, I'm thinking I'll wait a few weeks for the EK or DD blocks, and either return this or sell it cheap in the FS forum. When I do pull it off the card, I'll let everyone know what I find...

It would be interesting to know a few more temps, specifically if you could put the thermocouple between the board and the block, down between the inlet and outlet so that it’s just above the GPU (dodging any ram chips of course). That should give you a good idea of the temps of the block itself, which should give you an idea of where the problem lies. Also, I would see if you can get a good reading off of the inlet and outlet fittings themselves, and see what the water temp going in is and how much heat the water's pulling out.

Personally, I wish koolance wouldn’t gold plate the bottom of their blocks, since it actually hurts performance (in theory anyway). If I were you, I would try lapping the GPU area, you would probably have to cut a piece of glass that would fit, but I would bet that’s causing some of the problem.

One last thing, you might try putting a bead of gasket material down on the ridge that separates the two sides to insure that it’s not leaking (and that all the water is going through the GPU area, if that makes sense).
Dan

8800gtx 651/2100 EK block Idle 36c load 45c

Granted my cpu (Opty) is not dumping a lot of heat into the loop.

I'm waiting on a new mobo for my C2D and I'll post back with temps from that which I'm guessing will be higher

BobcatDan said:

It would be interesting to know a few more temps, specifically if you could put the thermocouple between the board and the block, down between the inlet and outlet so that it’s just above the GPU (dodging any ram chips of course). That should give you a good idea of the temps of the block itself, which should give you an idea of where the problem lies. Also, I would see if you can get a good reading off of the inlet and outlet fittings themselves, and see what the water temp going in is and how much heat the water's pulling out.

Personally, I wish koolance wouldn’t gold plate the bottom of their blocks, since it actually hurts performance (in theory anyway). If I were you, I would try lapping the GPU area, you would probably have to cut a piece of glass that would fit, but I would bet that’s causing some of the problem.

One last thing, you might try putting a bead of gasket material down on the ridge that separates the two sides to insure that it’s not leaking (and that all the water is going through the GPU area, if that makes sense).
Dan

Click to expand...

Agree 100% on your assessment. I believe most of the problem is the shoddy finish in the GPU area. The pictures don't show it very well, but there are pronounced ridges in that area from the mill. I also agree that the lack of a seal to the top plate from the center channel wall might be allowing liquid to pass over that area instead of through it.

I'm still debating on whether it's really worth the effort or if I should just drop it and grab an EK. I've already asked Alex at PTS if they can give me a heads up on the arrival of the EK GTS block.

Be back with some updates tonight or tomorrow...


Tom

So I was a bit disgusted by some of what was going on with this block and took a few days off from it. I decided I would lap the GPU contact area to smooth it out. However, I took it apart this morning to do just that, and it seems that surface or the GPU IHS is very far from even being square, let alone flat.

Pics:







I can still see the smoothing pattern in the area where the block didn't come in contact with the GPU, from when I spread the AS5 with a credit card (sample of course). I also think I probably had too much AS5 on there to begin with, but that sure doesn't make this any better.

This is something I'm not sure hand lapping will fix. What do you think? Opinions please...


Tom

Suck to see all the issues you are having with this block. I bought the GTX version since I couldn't find any other blocks in stock anywhere. This is for a new system build that I just finished, so right now I'm just running everything stock on air. This coming weekend I'll be installing all my wc parts, I'll let you know how I make out. Problem right now is I have no temperature monitoring on 64 bit Vista.

hey hemi
don't you have little bit too much of As5 there. i mean little bit less whould do the jop, and actually too much of it will hurt heat transfer,

That's a lapping job right there. Lapped my CPU this weekend, which was very far from flat and it too about 2 hours with 400 grit paper to flaten it (and it's about half the size fo the 8800 GPU and a quarter the size of the koolance block). If you start out with 200 it would speed things up, but that is a big task for sanding alone...
Dan