View Full Version : Swiftech stealth probs! Overheating!
KnownKiller
12-09-07, 04:47 AM
Hey guys.. I just spent all day and all night draining my water loop and removing my Switech VGA blocks and heatsinks from my 2 Evga 8800GTX's and installing the new blocks, Swiftech Stealth rev. 2. Low and behold i had probs getting good contact with the cores using the new blocks.
I put a long bead of AS 5 in the middle of the core and slapped the block on and tightened the block down and quickly checked it out for contact. What i seen i could'nt believe. I could see light threw the core and the block! I was like what the hell.. I shinned a flash light in the area and i could see the thero goop just hanging there inbetween. I quickly took the block off and noticed that the goop was spread a little bit in an oval shape but just in the very center of the core. Also it looked like the goop was not even flattened down at all meaning there was not much contact.
I first thought lets take a pic of this but i thought was lets just add more AS 5 and spread it out this time and try again.. I'll take pics of the next block.( note i didnt even notice the prob till i got the block on the second card ofcourse!)
I spread it out and put the block back on again and tighted the screws down all the way and checked it out and i didnt see much light as before and figured it was good. I did the same n the second block and no i didnt take pics :(
Put the cards in and took another 2 hours filling and bleeding the system.
Start the computer up and i'm like cool its alive( good sign). I get into winXp and check rivatuner for temps and what do i see??? 48/44 idle! ****ed off cause i had temps of 42/38 idle before with the old switech universal blocks with the G80 addapter.
I ran a bench of 3DMark 06 to see what the damage is on load and i ran the first test and hit escape and checked the temps and i got a 59/55! :bang head:( The highest i seen the temps with the old setup was 51/47.
This is with 1.6v GPU and 2.2 on the ram. Runing 684/1836/1000. <-- Not the highest i can go.
So it managed to complete the 3dmark 06 bench once.. the next time it rebooted the system.
So obviously i have a major prob on my hands right now.. Also it dont help that i'm just about out of AS5.
Now i've read that the newer rev stealth is bowed and its supposed to out do the older vid block that i have. What the hell do i do to fix this??
NOTE: the pads seem to thick for the memory chips as its bowing the card out.. i dont know is that is whats stopping the block from making good contact with the core of what. Looks like you would'nt need the dang pads as it seems like the bock would line up perfectly with all the memory chips and not bow out the card. Though i doubt it would have any pressure behind it to make a solid connect.
Not too happy right now as i blew $228 plus a 4 hour drive to get these locally so i could play this weekend when my 2 and a half year old is staying over her god mothers house this weekend.
Looks like i'm gona be spending the next 2 days putting back on all those dam sopper heatsinks back on the 2 GTX cards and the old blocks.
What a nightmare!
Oh yeah.. sorry its almost 6 am and no sleep!
Is there only one way to have the flow go with these blocks? I have the inlet going into the left side of the blocks and the out is the right. Would it make a difference it i reversed it? I mounted the lines that way as thats how i had it before and only way i could hook it up with my system without changing out the lines.
The only thing you can do is achieve proper contact. IMO you shouldn't have powered the system up until you were satisfied with the contact between the block.
Edit: to be honest I think those temps are more than fine. Not very hot if you ask me. It sucks that your temps went up, but it's not that bad. Just sell the new blocks and go back to what works better, or keep these knowing that those temps aren't dangerous. My card gets to over 85C sometimes. It's totally safe.
KnownKiller
12-09-07, 03:23 PM
heh they state that the blocks cool better then the MCW60 block and thats what i was expecting.. if there was going to be almost a 10C higher temps i would of went with the Danger Den full coverage blocks as my real only concern was cooling the rest of the card.
Dont know whatelse i could do to get a better contact as the screws are all tight. I wish i would of taken the block off again and checked the coverage before putting them back in but with those damn memory pads i didnt want to be out with nothing as i doubt you can remove the block more then twice without having to replace them.
Voodoo Rufus
12-09-07, 03:35 PM
1. Get some sleep.
2. After that, approach it with a fresh mindset. You say that the core is not making proper or adequate contact with the block, because of the memory pads. If that's true, try taking off the pads and see if the block will make proper contact with both the core and the memory. If not, you might try some different thermal pads.
A quick look on a popular cooling site yielded these: http://www.petrastechshop.com/ekrevoretimp.html
Might work.
KnownKiller
12-09-07, 11:08 PM
Welp finally bled the lines again of all air bubbles as that was pretty much my last idea.. and tested it out.
Same crap GPU1 and 2 - 48/45C Idle and 59/52C Load.
My last idea is that maybe since this block also touches the mem chips that since i'm runing 2.2v on it, maybe its also adding heat to the line.. i know the water isint touching the mem but the block is.
I'm gona flash the card to default mem speed and lower the voltage back down to 1.95 (default) and take a look.. I'll repoart back shortly.
KnownKiller
12-09-07, 11:11 PM
The only thing you can do is achieve proper contact. IMO you shouldn't have powered the system up until you were satisfied with the contact between the block.
Edit: to be honest I think those temps are more than fine. Not very hot if you ask me. It sucks that your temps went up, but it's not that bad. Just sell the new blocks and go back to what works better, or keep these knowing that those temps aren't dangerous. My card gets to over 85C sometimes. It's totally safe.
I thought the contact was fine.. its kinda hard to see with the block on.
I didnt change a thing from the lines, only the blocks. I expected what i read about the blocks, that they designed them to top the king (MCW60). I ended up seeing about a 10C incline to my temps.
KnownKiller
12-09-07, 11:50 PM
I just down clocked the mem speed and lowered the voltage on them from 2.20v to 1.95. Last ditch idea as i was thinking that maybe since the block covers the ram that with such high voltage it may also be heating the blocks up.
Results:
I got a whole -1C difference on idle temps. Dropped a little under load at 57/51.
So now i'm sittin at around a 6C increase using the new blocks.
Moto7451
12-10-07, 12:43 AM
Your card may have taller memory than most of the 8800s out there. Full cover blocks are sensitive to this. For example, the 6800 and 7800 have identical PCB layouts (pretty much identical PCBs really) but the newer memory on the 7800s is thinner so you need a new full cover block or use thick pads. Also, like in your case, the 7800 full cover blocks will not work on 6800s unless they're an oddball version with memory from 7800s.
Thats basically why I won't recommend a full cover block. I don't see spending an additional $100+ or $200+ (assuming an SLI system) every time you replace your video cards to be good for anyone except the water block companies.
KnownKiller
12-10-07, 01:29 AM
There both early 8800GTX models.. never heard anyone say anyting about any changes made to the 8800GTX cards... not saying that your wrong or anything just saying i didnt hear of a change.
Spawn-Inc
12-10-07, 03:00 AM
only think i can think of is take the block off and press down on the mem pads and or remove then all together and see if what kind of contact you get with mem and core. hopefully with the stupid pads gone you get nice contact on both!
KnownKiller
12-10-07, 03:16 AM
Yeah i have images in my mind of me takin the whole thing out again and checking without the pads on. i'd just hate to do that as if it dont make good contact i'm **** out of luck puttin the blocks back on.
It be nice if i could just return them but ya know since its watercooling i'm sure thats out of the question.
So possably in teh end i'm gona have to blow another $15? shipped for replacement pads and then loose my rear selling these blocks.
I'm happy that i choosed to have these blocks for my X-mas present from my parents. What a waste..
Voodoo Rufus
12-10-07, 09:05 AM
Another thing you can do is look for a local computer shop. They might have some cheap TIM or pads for sale. You don't HAVE to put AS5 or fancy TIM on memory.
KnownKiller
12-10-07, 11:34 AM
Welp i figure even after i get ahold of someone at swiftech there will tell me to take the blocks off and remount so i attempted to take the second block off whil still on the loop.. quickest way to see whats doing.
Pics taken while on my back as i had to hold the block up and take a shot underneith.
http://home.comcast.net/~eclypse/DSC07241.JPG
http://home.comcast.net/~eclypse/DSC07245.JPG
http://home.comcast.net/~eclypse/DSC07246.JPG
I've carfully taken the pads off and placed them on a peice of wax paper so i dont ruine the sticky back. I'm gona clean the goop off and do a test mount without the pads using goop on all the ram chips to see what kind of coverage i get without the pads this time.
Its gona be extremely ruff tryin to remount the block on my back but a mans gota do what a mans gota do.. Wish me luck.
evilrad
12-10-07, 01:27 PM
First of all, you used WAY too much AS5 there! Second, it looks like more pressure from the block is being placed toward the bottom of the core instead of the middle. I'd say that's your main problem.
KnownKiller
12-10-07, 01:34 PM
http://www.xtremesystems.org/forums/showthread.php?p=2617296#post2617296
This is why I refuse to buy or use full cover water blocks.
KnownKiller
12-10-07, 03:04 PM
This is why I refuse to buy or use full cover water blocks.
I feel ya man..
KnownKiller
12-10-07, 03:05 PM
Here we go...
First a shot of the idle temps before with the older swiftech MCW60 blocks in SLI.
http://home.comcast.net/~eclypse/3700MHZ-684-1836-1000-winxp-YES!.jpg
Now a pic of the idle temps after installing the 2 Stealth blocks in SLI.
http://home.comcast.net/~eclypse/3700MHZ-684-1836-1053-winxp-2.05vmem.jpg
Heres a shot of the card removed.
http://home.comcast.net/~eclypse/DSC07254.JPG
Heres a pic showin the new AS Ceramique spread over everything.
http://home.comcast.net/~eclypse/DSC07257.JPG
Close up.. I spread the goop with a credit card as usual.
http://home.comcast.net/~eclypse/DSC07259.JPG
A pic of the card now installed.
http://home.comcast.net/~eclypse/DSC07261.JPG
Close up pic so you can see the AS goop on the mem chips compared to the first block with the pads.
http://home.comcast.net/~eclypse/DSC07262.JPG
Just a couple of pics showin the overall cooling of the case.
http://home.comcast.net/~eclypse/DSC07268.JPG
http://home.comcast.net/~eclypse/DSC07269.JPG
http://home.comcast.net/~eclypse/DSC07270.JPG
KnownKiller
12-10-07, 03:06 PM
http://home.comcast.net/~eclypse/DSC07272.JPG
Now the good stuff..
Idle temps
http://home.comcast.net/~eclypse/temps-1CardNopads.JPG
And now load!
http://home.comcast.net/~eclypse/LOAD temps-1CardNopads.JPG
Now i dont know whats up with the system now though.. The mem look like there making fine contact, ofcrouse with the trouble of getting that card on the block in the position that i did (Miracle really) i didnt want to chance taken it off and puttin it back on again.
So i made it through 2 loops of 3dmark 06 and i seen that it finished the last loop as i past by the system coming in from outside. I come back 2 mins later and i see that the system rebooted :(
Then while the system was idling and i was on the notebook trying to upload pics the system blue screened on me. Forgot the error ofcourse but at the bottom it said starting physical memroy dump. I turned it off.
So i thought my current prob would be if i wanted to dump the lines and pull both cards out and redue them without pads to now wondering it something is cookin like the mem with the pads off. I did leave the little blue pad on the shadder chip as i didnt feel like chipping it putting the block on in the position i did. I think that may be preventing the memory on that side near the shadder chip from having good contact.
Those blocks are pretty heavy, are you sure your graphics cards aren't getting torqued a bit in their slots and loosing contacts with the PCI-E bus?
KnownKiller
12-10-07, 03:52 PM
No there not heavy. There made out of Aluminum with the base plate for the core is copper. I just checked under the motherboard as i can easily see under there with a flash light and its not bowing or anything. I believe the card slot bracket would help hold the weight anyways.
No there not heavy. There made out of Aluminum with the base plate for the core is copper. I just checked under the motherboard as i can easily see under there with a flash light and its not bowing or anything. I believe the card slot bracket would help hold the weight anyways.
Well, it was worth a shot.
NGL_BrSH
12-12-07, 02:39 PM
dood.. WAY too much arctic silver on both applications.. it's not putty.. it's to fill microscopic gaps not one-eighth inch gaps.
If i were you, i'd first make sure both surfaces were truly flat. (ie, waterblocks and gpu). You could us a metal right-angle ruler or a something else that is hard and perfectly flat like glass.
If both surfaces are flat, then it wouldn't make sense that the two surfaces wouldn't mate up properly. That would lead me to believe the block isn't being tightend properly and/or seated properly. Could be touching something making it sit higher on one side than the other..
If the surfaces aren't both flat, you should either rma the block or lap both surfaces.
Then finally this is not Head & Shoulders.. do NOT apply liberally. A very thin line of and a little twist 90 degrees and back should do it.. by applying too much you will actually hinder the rate of which heat may be transfered between the block and the gpu.
anyway, gl, if it were me i'd lap it anyway or at least lap the waterblocks.
Thing is, your temps look fine.
My questions:
1) Have you changed anything else. Your over clock, voltages? etc? (cpu or gpu)
2) Have you checked to see if you may have had a leak.
KnownKiller
12-12-07, 03:32 PM
dood.. WAY too much arctic silver on both applications.. it's not putty.. it's to fill microscopic gaps not one-eighth inch gaps.
If i were you, i'd first make sure both surfaces were truly flat. (ie, waterblocks and gpu). You could us a metal right-angle ruler or a something else that is hard and perfectly flat like glass.
If both surfaces are flat, then it wouldn't make sense that the two surfaces wouldn't mate up properly. That would lead me to believe the block isn't being tightend properly and/or seated properly. Could be touching something making it sit higher on one side than the other..
If the surfaces aren't both flat, you should either rma the block or lap both surfaces.
Then finally this is not Head & Shoulders.. do NOT apply liberally. A very thin line of and a little twist 90 degrees and back should do it.. by applying too much you will actually hinder the rate of which heat may be transfered between the block and the gpu.
anyway, gl, if it were me i'd lap it anyway or at least lap the waterblocks.
I put that much to make sure that something touches.. besides in the end the excess will squeeze out with pressure. Gabe (CEO of Swiftech) siad i had the best contact i could hope for after seeing the imprint left on the block after removal. The blocks are purposely bowed (concaved) for better heat transfer.
In the end the block was bending the end of my card outward pretty bad and Gabe though i'd should RMA the block with Siftech and give another a try just encase the block was off.
I'm starting to think that maybe my TIM on the GPU core is not flat and i should'nt use a full cover block.
I'll no more when i take both blocks off today and try to fix the first one on the second card.. if it does the same bending then theres no hope for this block on my card.
There sending 2 new blocks just encase its just a bad batch.
KnownKiller
12-12-07, 03:34 PM
Thing is, your temps look fine.
My questions:
1) Have you changed anything else. Your over clock, voltages? etc? (cpu or gpu)
2) Have you checked to see if you may have had a leak.
I've only changed the blocks and nothing else.. not even a line. just swaped out the blocks.
Check the thread at xtremesystems.org (http://www.xtremesystems.org/forums/showthread.php?t=168764&page=2)as theres more pics and info on the bending. Post 42 and below.
thideras
12-12-07, 03:35 PM
dood.. WAY too much arctic silver on both applications.. it's not putty.. it's to fill microscopic gaps not one-eighth inch gaps.
If i were you, i'd first make sure both surfaces were truly flat. (ie, waterblocks and gpu). You could us a metal right-angle ruler or a something else that is hard and perfectly flat like glass.
If both surfaces are flat, then it wouldn't make sense that the two surfaces wouldn't mate up properly. That would lead me to believe the block isn't being tightend properly and/or seated properly. Could be touching something making it sit higher on one side than the other..
If the surfaces aren't both flat, you should either rma the block or lap both surfaces.
Then finally this is not Head & Shoulders.. do NOT apply liberally. A very thin line of and a little twist 90 degrees and back should do it.. by applying too much you will actually hinder the rate of which heat may be transfered between the block and the gpu.
anyway, gl, if it were me i'd lap it anyway or at least lap the waterblocks.He nailed it. You used WAY too much thermal paste on the core and especially the memory. The memory chips are not cake and the paste is not icing.
I put that much to make sure that something touches.. besides in the end the excess will squeeze out with pressure. Gabe (CEO of Swiftech) siad i had the best contact i could hope for after seeing the imprint left on the block after removal. The blocks are purposely bowed (concaved) for better heat transfer.
In the end the block was bending the end of my card outward pretty bad and Gabe though i'd should RMA the block with Siftech and give another a try just encase the block was off.
I'm starting to think that maybe my TIM on the GPU core is not flat and i should'nt use a full cover block.
I'll no more when i take both blocks off today and try to fix the first one on the second card.. if it does the same bending then theres no hope for this block on my card.
There sending 2 new blocks just encase its just a bad batch.
Tough break man.
It's just WAY hard to machine a full coverage block for a graphics card, considering all the different chips the different manufactures use that, that will be foolproof.
NGL_BrSH
12-12-07, 04:11 PM
I put that much to make sure that something touches.. besides in the end the excess will squeeze out with pressure. Gabe (CEO of Swiftech) siad i had the best contact i could hope for after seeing the imprint left on the block after removal. The blocks are purposely bowed (concaved) for better heat transfer.
Haha.. WHAT?! concave for better heat transfer? they just need to meet properly.. they both shouldn't be concave.. swifttech makes the waterblock convex because most chips aren't flat.. they are concave.. so that they mate tight. Space is created both being concave.. space is filled with air. Air is a good thermal insulator.. not what you want.. best is to just make them both flat.
If you squeeze a ton of "goop" on there and squeeze them together you will still have a ton of "goop" left between.. it won't press out the excess to a point where you will be left with a perfect layer. you'll be left with a arctic -silver sandwich.
KnownKiller
12-12-07, 04:20 PM
Haha.. WHAT?! concave for better heat transfer? they just need to meet properly.. they both shouldn't be concave.. swifttech makes the waterblock convex because most chips aren't flat.. they are concave.. so that they mate tight. Space is created both being concave.. space is filled with air. Air is a good thermal insulator.. not what you want.. best is to just make them both flat.
If you squeeze a ton of "goop" on there and squeeze them together you will still have a ton of "goop" left between.. it won't press out the excess to a point where you will be left with a perfect layer. you'll be left with a arctic -silver sandwich.
Yes maybe with AS 5 as its thicker and made out of silver but using ASC made the difference and it was able to squeeze out like it should.
Concaved block means only the dead center will touch and thats where it made contact as you can tell in the pic.
http://home.comcast.net/~eclypse/DSC07274.JPG
thideras
12-12-07, 04:23 PM
Actually...that looks like it made no contact...unless you put the goop in the center....then it pushed it all out. STILL, you are using WAY too much paste.
Also, in the center, there is a dark circle...looks like the GPU is burnt :confused:
Yes maybe with AS 5 as its thicker and made out of silver but using ASC made the difference and it was able to squeeze out like it should.
Concaved block means only the dead center will touch and thats where it made contact as you can tell in the pic.
http://home.comcast.net/~eclypse/DSC07274.JPG
Woldn't that be Convex? Anyways... Thats still a butload of paste.
I'm not seeing the reasoning behind having a non flat surface. Isn't the idea to have maximum contact, not minimal? Otherwise you are using gobs of thermal paste just to transfer heat, rather than a fine film to produce even transfer.
You sure these aren't just a bad batch of blocks?
KnownKiller
12-12-07, 05:41 PM
Actually...that looks like it made no contact...unless you put the goop in the center....then it pushed it all out. STILL, you are using WAY too much paste.
Also, in the center, there is a dark circle...looks like the GPU is burnt :confused:
Thats the way the new blocks (Stealth and GTX) are as being bowed in the center.. So only the center area will touch the core. Thats what Gabe the CEO of Swiftech told me as what you want to see, just the center area touching. Its been tested and gives 1-2 C better temps.
:confused: How in tar-nation is that supposed to work?
KnownKiller
12-12-07, 05:48 PM
Woldn't that be Convex? Anyways... Thats still a butload of paste.
I'm not seeing the reasoning behind having a non flat surface. Isn't the idea to have maximum contact, not minimal? Otherwise you are using gobs of thermal paste just to transfer heat, rather than a fine film to produce even transfer.
You sure these aren't just a bad batch of blocks?
I'm not sure if there bad or not but Gabe said the bent card should'nt happen so i guess there made wrong or my card is off.
The temps are totally fine in the end after putting the card back on and tighting the 4 GPU screws first then the 2 for the NVIO chip last.
I just got the first card off and the contact so just a tad better then the 2nd card. But i tightened all the screws together and nott eh core first then the NVIO last.. so it could of been better and it could of ended up bending that card as well if i did it right. So it may be a bad batch. The store i bought it at told me the've never had a RMA on a Stealth block before so.. Gabe said he's never seen this prob after selling thousands of these blocks. COuld be a bad batch or my cards are off.
I had 1.6v on the cards and they were overlcocked to 684/1836/1000, so the temps were just about on with what i was seeing with the MCW60 blocks after i got the 2nd card on good.
I also tested the cards at default clocks and voltage and the temps were under 38-40C idle.
KnownKiller
12-12-07, 05:48 PM
:confused: How in tar-nation is that supposed to work?
No idea but thats the change with the 2nd version of the blocks.
That whole convex thing just sounds fishy to me. Seems like you would bend the card liek you said. Personally I would send em back, and get my money back.
NGL_BrSH
12-12-07, 06:55 PM
No. They really are convex from swiftech.. but it's because most of the gpu's are coming concave... Lap both and you'll have the perfect match. they will both be flat. Gpu's typically are concave because manufacturing isn't perfect but still are within spec. Spend 1 hour lapping both your gpu and your waterblock both flat and put a very thin layer of ceramique and i bet you would enjoy your results.
When i lapped my cpu and apogee block i enjoyed a drop in temps in avg of 4-5c with a much smaller dynamic range. Took me an hour and it was free.
KnownKiller
12-12-07, 07:05 PM
I think i may of figured out the prob now after lookin at both blocks after i got them out of the case and loop.
It seems that the allen screws that hold that GPU plate to the block were very tight.. so tight that the end of the screws are pretty much flush with the GPU copper plate.
Now on the 2nd block, (the one that was making the card bend) it looked like all 4 bolts were not tightened.. Look like they could use 2-3 turns. So in the end it looks like because of the GPU copper plate is not totally seated and tight to the block, it may be why the thing is crooked and causing the block to bend the card when totally tightened on.
I wish i had the right size allen key for the 4 bolts to try this.. But i guess Swiftech will notice this when they get the blocks back.
I just got notice that the package has been sent by UPS ground. So it will be awhile before i get the new blocks.
Gabe asked me about this and i thought he was talking about the standoffs where the screws go in when you put the card on the block.
KnownKiller
12-12-07, 07:08 PM
No. They really are convex from swiftech.. but it's because most of the gpu's are coming concave... Lap both and you'll have the perfect match. they will both be flat. Gpu's typically are concave because manufacturing isn't perfect but still are within spec. Spend 1 hour lapping both your gpu and your waterblock both flat and put a very thin layer of ceramique and i bet you would enjoy your results.
When i lapped my cpu and apogee block i enjoyed a drop in temps in avg of 4-5c with a much smaller dynamic range. Took me an hour and it was free.
I could just see what the pressure would be on the memory chips after i lapped both the card and the blocks. its already bowing the card to where i'm not comfortable.
I'd try that on a non full coverage block but i still dont want to lap my 8800GTX cards that cost me $650 each way back when they first came out hehe.
Spawn-Inc
12-12-07, 07:28 PM
if you have two water blocks that have the same amount of pressure but one is flat and the other is bowed (convex) the bowed one will cool better. the reason is because the extra pressure right on the core helps pull the heat out better then the flat block.
if you have two water blocks that have the same amount of pressure but one is flat and the other is bowed (convex) the bowed one will cool better. the reason is because the extra pressure right on the core helps pull the heat out better then the flat block.
Sounds more like a manaufacturing cop out in order to make up for the sloppy aplication of TIM on the part of most users, and sloppy seating of blocks.
just my $0.02 :shrug:
Moto7451
12-13-07, 01:05 AM
Personally I wouldn't buy a bowed GPU block since future cards will not have the IHS. The new G92 based cards for example do not. The whole point of the bow is to counter a convex IHS. The actual die itself though is flat.
KnownKiller
12-13-07, 01:50 AM
Personally I wouldn't buy a bowed GPU block since future cards will not have the IHS. The new G92 based cards for example do not. The whole point of the bow is to counter a convex IHS. The actual die itself though is flat.
Wow thats a good thing forsure! I guess they finally learned that the core is way cooler without it on there.
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.