Temperature/Block Comparison For 8800GTX

Printable View

03-20-2007, 08:39 PM
Snyxxx

5 Attachment(s)

Temperature/Block Comparison For 8800GTX

I have done a little testing to convince myself what the best cooling solution for this beast of a video card. Used Everest Ultimate for all temperature
mesurements and the AC is already on for the year so ambient is pretty constant 72°F (~20°C).

Multi-core Prime 95 25.2 (Small FFTs) used for heating CPU and ATI Tool used for heating GPU. I found OCCT does not produce as much heat as above. I lapped the Xeon 3070 for peace of mind that the concavity is gone. All CPU temps are at the core (max) per Everest.

Watercooling CPU Only With Stock Air Cooled 8800GTX:

http://www.xtremesystems.org/forums/...1&d=1174450293

CPU Idle: 24°C
CPU Load: 34°C

8800GTX Idle: 63°C (Diode)
8800GTX Load: 84°C (Diode)

Observations - 8800GTX Temps are a little disturbing. Happy with CPU temps.

Watercooling CPU And 8800GTX Using MCW60 With Heatsinks:

http://www.xtremesystems.org/forums/...1&d=1174450656
http://www.xtremesystems.org/forums/...1&d=1174450693

CPU Idle: 27°C
CPU Load: 39°C

8800GTX Idle: 36°C (Diode)
8800GTX Load: 44°C (Diode)

Observations - I hated dealing with all those heat sinks. The heatsink on the lone naked chip that looks like a NB or SB got so hot I could not put my finger on it more than half second. if I were to keep this option, a bigger heatsink would be required or a fan. The GPU (ambient) got to 51°C. Maybe it was from this little chip. All the ram heatsinks were cool to the touch. The VRM heatsinks were warm, but not too bad.

Watercooling CPU And 8800GTX Using EK Full Cover Block:

http://www.xtremesystems.org/forums/...1&d=1174451064

http://www.xtremesystems.org/forums/...1&d=1174451099

CPU Idle: 30°C
CPU Load: 41°C

8800GTX Idle: 39°C (Diode)
8800GTX Load: 48°C (Diode)

Observations - Results show a little more heat than the GPU only solution above. I guess some of the VRM, Memory and little chip's heat is getting into the water. 3-4°C difference is not that bad. Additional flow resistance may not help either. The GPU (ambient) temperature so hot above was only 40°C for this case. I felt all over the block and felt no hot spots. I like the tube routing with the EK block better as well.

Conclusion:

I knew I was going to use the EK full cover block in the end, but always wanted to know what the difference would be. These types of things bug me for not knowing.

I already had a MCW60 laying around and just had to buy the G80 bracket. I kept everything the same as much as possible. I believe that just about any water cooling solution used for the 8800GTX would be vastly better than the stock air cooler. Temps were almost cut in half.

By the way the 8800GTX at full load added about 7-8°C the the CPU core temperature.

Hope this helps someone.
03-20-2007, 08:40 PM
serialk11r

I never thought a fullcover block would hurt flow that much...
03-20-2007, 08:54 PM
AuroraProject

How much did the EK increase the cpu temps at idle?
03-20-2007, 09:12 PM
Omastar

Glad I got that MCW60 now. Presents me with an upgrade path and better flow. I don't really care about having to deal with 12 heatsinks for the RAM/PWM/NVIO. And for me, the NVIO chip on my GTS never got hot with an MC14, even under ATI_Tool.
03-21-2007, 01:58 AM
Eddy_EK

I am sure all would be laughing if you added a DangerDen full cover block :D
03-21-2007, 06:00 AM
Snyxxx

From what I have read, the DD full cover block would be worse. Don't get me wrong, I love the EK block and planned on using it from the beginning. That is why I bought two. I personally did not like the MCW60 option since it only cooled the core, made my tubing runs more difficult and the other hot points got real hot.

Also, I am not sure if the temps increased from flow resistance or the summation of individual heat sources being added to the water. By using a Storm, I consider flow restriction a minor point.

To me, it does not matter. I consider the EK block a work of art and prefer using it.
03-21-2007, 07:46 AM
IanY

Quote:

Originally Posted by Snyxxx

To me, it does not matter. I consider the EK block a work of art and prefer using it.

Are you building art or are you building a watercooled computer? It seems like you are basing your decisions largely on appearances.
03-21-2007, 08:38 AM
lasse.j5

Good job mate.

What does that NVIO chip actually do? Has is anything to do with Overclocking results?
03-21-2007, 09:20 AM
Snyxxx

Ian Y,

I like to have both appearances and functionality.:D

If I can have appearances for a few degrees, so be it. Its just a hobby, but I am a little particular in the looks of things.

lasse.j5,

Thanks. I do not know what that chip does, but it is hot, hot, hot. I have not tried any overclocking yet. Trying to establish a base line currently.
03-21-2007, 11:19 AM
masaville

Thanks for doing the test. Very useful info!!
03-21-2007, 11:19 AM
masaville

Quote:

Originally Posted by Eddy_EK

I am sure all would be laughing if you added a DangerDen full cover block :D

I suspect my DD blocks are worse. My temps are slightly higher, but it's negligible for the most part. We're talking a few degrees at most. EK blocks seem to be machined better though from the pics I've seen.
03-21-2007, 12:22 PM
IanY

Quote:

Originally Posted by ranker

Thanks for doing the test. I just put an order in for Swiftech's MCW60's at Petra's myself. I'd rather have the raw performance even though the EK's are much more attractive. The only thing I'll miss is the ease of setting up an SLI setup without having to bend the tubes full round.

Does anyone know if the 8900 GTX's will require a different mounting addon than he one used for the 8800's for the MCW60's? I'm planning on using the step up program and don't want these waterblocks to go to waste.

OOOooo... $10 million question!!! I'm personally waiting to see the 8900 cards before buying more video blocks hehehe :)

Yes.. I want 8900 cards.. My 7900s were so much better than my 7800s.. so I expect the 8900s to be cooler running and have a higher overall product quality.
03-21-2007, 12:22 PM
IanY

Quote:

Originally Posted by masaville

I suspect my DD blocks are worse. My temps are slightly higher, but it's negligible for the most part. We're talking a few degrees at most. EK blocks seem to be machined better though from the pics I've seen.

I love the smack talking.
03-21-2007, 12:39 PM
ranker

Quote:

Originally Posted by IanY

OOOooo... $10 million question!!! I'm personally waiting to see the 8900 cards before buying more video blocks hehehe :)

Yes.. I want 8900 cards.. My 7900s were so much better than my 7800s.. so I expect the 8900s to be cooler running and have a higher overall product quality.

I agree. I only bought the 8800GTX's so that I could either RMA them for the 8900 or use the step up program. My 7900's were great at a much lesser cost than my 7800 SLI's. I'm hoping it stays the same pattern with more performance/features at a 10-20% less cost.

My only worry is that the 8900's won't be compatible with the MCW60's. I just ebayed off my EK covers and recouped 80% of my investment so I'm feeling good for now.

Let's hope the 8800 mounting kit is all we need. I'd hate to get nickled and dimed more just to run the 8900's.
03-21-2007, 01:58 PM
serialk11r

Hmmmm... I like the non-fullcover look more. Anyways, there is no way that small chip you are talking about plus the ram and everything will increase water temps by that much. It has to be the restriction.
03-21-2007, 02:26 PM
[cTx] Nooc

jesus I'm using the EK block on my 8800 because only about 100% of you suggested it over mcw60. now it's 50/50...oh well, I don't regret it! beautiful block, and comes with chrome barbs (deal breaker right there).
03-21-2007, 02:28 PM
serialk11r

Quote:

Originally Posted by Nooc

jesus I'm using the EK block on my 8800 because only about 100% of you suggested it over mcw60. now it's 50/50...oh well, I don't regret it! beautiful block, and comes with chrome barbs (deal breaker right there).

Lol chrome barbs... If Swiftech tapped all their stuff to 1/4" BSPP, that would be cool. It was the reason I got my Coolrad 22T not the swiftech thingy, because it had metal barbs :) Performance (heat dissipated/flow) is something like either a few percent better or worse. Of course, when you can't afford to spend an extra 20 bucks on metal barbs (i.e. me) then plastic is the way to go.
03-21-2007, 02:43 PM
masaville

Quote:

Originally Posted by IanY

OOOooo... $10 million question!!! I'm personally waiting to see the 8900 cards before buying more video blocks hehehe :)

Yes.. I want 8900 cards.. My 7900s were so much better than my 7800s.. so I expect the 8900s to be cooler running and have a higher overall product quality.

Ian, I'm sure the 8900s will actually decrease performance. Nobody told you???? haha j/k :stick:
03-21-2007, 02:44 PM
masaville

Quote:

Originally Posted by serialk11r

Hmmmm... I like the non-fullcover look more. Anyways, there is no way that small chip you are talking about plus the ram and everything will increase water temps by that much. It has to be the restriction.

It's the sister of the SB MCP on a 680i board. =P Super hot too!! :D
03-21-2007, 02:46 PM
serialk11r

Quote:

Originally Posted by masaville

It's the sister of the SB MCP on a 680i board. =P Super hot too!! :D

Hot is one thing, heat is another. The chip gets hot because the cooling is not great. You can have a great deal of heat and cool the heat source down, and you can have something that doesn't put out much heat with bad cooling, and get it very very hot. I'm sure those mobile processors that draw very very little power would get hot with only an MC14 ramsink on it lol.
03-21-2007, 02:49 PM
pplapeu

looks good for EK

I might be the only Innovatek full card guy you hear from. Mine has been on for a week. This on a XFX-XXX GTS 580 clocked card.

OK.....before I shoot out a load number, realize i have a huge external cooler. I run a 2g res, two cores and two inline rads with a 5gpm pump and run only distilled water.

I live in a cold latitude and lately the room temp is 70F.

41C seems to be the load temp I see after an hour or so of gaming and that number is not varing much yet.....so see you in summer!

I bet you hate advice. BUT HEY! .... here you go. Get some more water. Inline a real gallon of extra juice outside the case [forget LAN parties unless you use quick connect] and inline a heater core or two. I predict the block does better.:stick:

PS.....might be getting EK if they become available here and the price stays were they are. My Innovatek cost beaucoups.
03-21-2007, 02:55 PM
Snyxxx

Nooc,

Look at my picture of the MCW60. I just used DD High Flow barbs. Also, on the reservoir. The NPSM threads work on G1/4 barbs since the threaded part is very short.
03-21-2007, 02:57 PM
serialk11r

Quote:

Originally Posted by Snyxxx

Nooc,

Look at my picture of the MCW60. I just used DD High Flow barbs. Also, on the reservoir. The NPSM threads work on G1/4 barbs since the threaded part is very short.

Yea, people use the DD barbs on Swiftech stuff, but I wouldn't because they weren't meant to.
03-21-2007, 03:08 PM
pplapeu

FORGOT TO SIGGIE IN MY RIG!!!!
Processor
Model : Intel(R) Pentium(R) 4 CPU 3.20GHz
Speed : 3.21GHz-------Mod # SL7AA
Performance Rating : PR4271 (estimated)
L2 On-board Cache 512kB L3 On-board Cache 2MB
Asus P4GPL-X -- Total Memory : Kingston KHX3200AK2/2G
Floppy disk drive
Hard Disk : WDC WD740GD-00FLC0 [69.2GB]
PIONEER DVD-117R
SONY CD-RW CRX175E
Microsoft Windows XP/2002 Home (Win32 x86) 5.01.2600 (Service Pack 2)
Marvell Yukon 88E8053 PCI-E Gigabit Ethernet Controller
D-Link 4100 ["GameFuel" enabled}
Monitor/Panel : Plug and Play Monitor Dell 24"
XFX-XXX PV-T80G-GHD9 Cool-Matic nVidia 8800
Innovatek Mod# 501254
Aspire ATX-AS680W-BK
Logitech G15 -- Razer Diamondback -- JBL Creature
03-21-2007, 03:10 PM
Fairydust

Quote:

Originally Posted by serialk11r

Hmmmm... I like the non-fullcover look more. Anyways, there is no way that small chip you are talking about plus the ram and everything will increase water temps by that much. It has to be the restriction.

Without actually measuring water temp that is a bold statement. Whats the efficiency of the voltage regulation? Power consumption of ram? What fans at what voltage were used?

Pa 120.2 with Nexus at 7 V would net a 2,5 Degree water temp increase for ~35 Watts additional heat dump, 1.
03-21-2007, 03:16 PM
serialk11r

Quote:

Originally Posted by Fairydust

Without actually measuring water temp that is a bold statement. Whats the efficiency of the voltage regulation? Power consumption of ram? What fans at what voltage were used?

Pa 120.2 with Nexus at 7 V would net a 2,5 Degree water temp increase for ~35 Watts additional heat dump, 1.

Now where does it say that???????? WTF are you talking about? I don't see anything that compares water/air temp difference to heat dissipated. I only see "10C coolant/air difference".
You think ram consumes lots of power? At 1GPM, it takes 250W of heat to get the water up 1C. Of course, the rad won't dissipate all of it. If the ram and stuff really was that hot, then I don't think ramsinks would do the job...
03-21-2007, 03:27 PM
Vapor

Quote:

Originally Posted by serialk11r

At 1GPM, it takes 250W of heat to get the water up 1C.

That forumula is for coolant (well, water) temperature differences immediately before and after the heatsource. Not a sustained load coolant temp. If that were true, radiators would be useless.

Considering the 8800GTX pulls ~150W (being nice) when OC'd and the core uses at most 110W of that (closer to 100W if I had to guess), and say the EK absorbs 2/3rds of it, that's another 25-35W of heat, and on a 120x2 radiator, that could easily lead to the few C increase for coolant temps for the whole loop (witnessed in both GPU and CPU temps). Restriction is also playing a part, but the effects of restriction are probably way overrated.
03-21-2007, 03:29 PM
Fairydust

Quote:

Originally Posted by serialk11r

Now where does it say that???????? WTF are you talking about? I don't see anything that compares water/air temp difference to heat dissipated. I only see "10C coolant/air difference".
You think ram consumes lots of power? At 1GPM, it takes 250W of heat to get the water up 1C. Of course, the rad won't dissipate all of it. If the ram and stuff really was that hot, then I don't think ramsinks would do the job...

Check the thermal resistance vs flow rate for PA120.2 chart, 0.072 (°C/W) * 35 = 2,52 °C.
03-21-2007, 03:30 PM
serialk11r

Quote:

Originally Posted by Vapor

That forumula is for coolant (well, water) temperature differences immediately before and after the heatsource. Not a sustained load coolant temp. If that were true, radiators would be useless.

Considering the 8800GTX pulls ~150W (being nice) when OC'd and the core uses at most 110W of that (closer to 100W if I had to guess), that's another 40-50W of heat, and on a 120x2 radiator that's already stressed, that easily leads to the few C increase for coolant temps for the whole loop (witnessed in both GPU and CPU temps).

Yes, but I'm sure there are other components that use some energy as well. You've seen the fullcover blocks' base designs right? The water doesn't flow over some components, so the thermal resistance should be pretty high. If the core runs hotter, then less heat is making it into the loop as well right? That's another thing ;)
03-21-2007, 03:32 PM
serialk11r

Quote:

Originally Posted by Fairydust

Check the thermal resistance vs flow rate for PA120.2 chart, 0.072 (°C/W) * 35 = 2,52 °C.

The thermal resistance is dependent on coolant/air temp difference lol.
03-21-2007, 04:39 PM
AuroraProject

Quote:

Originally Posted by AuroraProject

How much did the EK increase the cpu temps at idle?

Anyone?
03-21-2007, 04:49 PM
pplapeu

Way to conservative a number

"Considering the 8800GTX pulls ~150W (being nice) when OC'd and the core uses at most 110W "

I would say 150w for the core alone. It is practically an order of magnitude larger that any CPU I ever laid eyes on, and I read it has more transistors than any CPU ever.

Stock cooling people report 65C idling and 80+C loaded. My GTS showed up nearly as hot with the same chip dumbed down a tad.
03-21-2007, 06:55 PM
serialk11r

Quote:

Originally Posted by pplapeu

"Considering the 8800GTX pulls ~150W (being nice) when OC'd and the core uses at most 110W "

I would say 150w for the core alone. It is practically an order of magnitude larger that any CPU I ever laid eyes on, and I read it has more transistors than any CPU ever.

Stock cooling people report 65C idling and 80+C loaded. My GTS showed up nearly as hot with the same chip dumbed down a tad.

What makes you think that the core will pull 150W? Just because it runs hot doesn't mean its generating that much heat. A soldering iron might only have 20W going through it but it can reach 250C or more.
03-21-2007, 06:57 PM
serialk11r

Quote:

Originally Posted by AuroraProject

Anyone?

Read: It said with the EK block temps on CPU were 30 idle, and with only the CPU in the loop, it was 24C ;)
03-21-2007, 08:42 PM
Vapor

Quote:

Originally Posted by serialk11r

Yes, but I'm sure there are other components that use some energy as well. You've seen the fullcover blocks' base designs right? The water doesn't flow over some components, so the thermal resistance should be pretty high.

Thermal resistance may be kinda high, but the whole point and DESIGN of the block is to draw heat away from all the hot components and into the water. Just because it's not OUTSTANDING at it doesn't mean it's not doing it. The heat is getting to the coolant. Witness the GPU ambient temps dropping from 51C (MCW60) to 40C (EK)...

Quote:

If the core runs hotter, then less heat is making it into the loop as well right? That's another thing ;)

Right, except CPU temps went up as well, which is solely indicative of either coolant or ambient temp rising....
03-21-2007, 08:45 PM
Vapor

Quote:

Originally Posted by pplapeu

"Considering the 8800GTX pulls ~150W (being nice) when OC'd and the core uses at most 110W "

I would say 150w for the core alone. It is practically an order of magnitude larger that any CPU I ever laid eyes on, and I read it has more transistors than any CPU ever.

Stock cooling people report 65C idling and 80+C loaded. My GTS showed up nearly as hot with the same chip dumbed down a tad.

ATI pegged the R580 core drawing only a little over half the card's power at stock voltages. So I am being generous in my argument when I say 110W out of 150W.

0.072 (°C/W) * X = 5 °C, X is equal to 70W, so 105W (70 + 35 quoted earlier) may be the total card draw at these clocks. 5C delta comes from CPU temp increase. Either way, the added heat dump from the EK's full-cover is 50% higher than just the core alone based on dT, so 100W out of 150W when OC'd is probably not far fetched either.
03-22-2007, 12:15 AM
Krowten-Dog

Does the fact that you did not use a washer/o-ring spacer under the blank plug on the copper side make a difference?

I noticed on my (trusty calipers) block that without the spacer under the plug, that the plug extended ~1.5 mm into the 5 mm water channel. Would that loss of 1.5 mm make any difference in waterflow?
03-22-2007, 12:37 AM
Whokidmo

Question, does air cooling the mem chips on the VGA and watercooling the main chip decrease your max OC, i mean the mem chips might over heat etc, hey OP can you do some major OCing? i honestly think air cooling and WC don't mix and that even if u do push the 2 set ups to the same speeds, 700/1200, you'll eventually burn out the swiftech card cause the mem chips get so hot with you OC, thats just my 2 bits at the end of the day, WCing is better then ACing which is why were all upgrading so why use a :banana::banana::banana::banana:ty hybird? sure the main chip temp is down but i honestly i'm like 99.9999% sure that the temp on the mem chips will increase by alot with the swiftech block or the mem chips will burn out, SHOW SOME OC TESTS.

btw i remind people many said that southbridges and vregs on mobos didn't need to by WCed... and now many disagree.
03-22-2007, 02:21 AM
Ambro

Quote:

Originally Posted by Krowten-Dog

Does the fact that you did not use a washer/o-ring spacer under the blank plug on the copper side make a difference?

I noticed on my (trusty calipers) block that without the spacer under the plug, that the plug extended ~1.5 mm into the 5 mm water channel. Would that loss of 1.5 mm make any difference in waterflow?

Good question - I didn't find anything in Eddy's instructions about the use of those two spacers.

Eddy can you shed some light on how we are supposed to hook these in?
03-22-2007, 04:09 AM
Snyxxx

First, thanks to everyone for the compliments.

Second, actually I did think about the spacers on the EK block. I read somewhere on these forums that they keep the threaded area of the barb from protruding into the flow channel.

Before I screwed the barbs in, it was obvious that the barb screwing into the copper side would protrude into the water channel. However, the delrin side was a different story. The threads were not as thick as the delrin piece and figured I did not need the spacer.

In hind sight, I probably should have just put the other spacer in and been done with it.
03-22-2007, 02:04 PM
pplapeu

my reason...more x-sistors = more heat

Quote:

Originally Posted by serialk11r

What makes you think that the core will pull 150W? Just because it runs hot doesn't mean its generating that much heat. A soldering iron might only have 20W going through it but it can reach 250C or more.

It is practically an order of magnitude larger that any CPU I ever laid eyes on, and I read it has more transistors than any CPU ever
03-22-2007, 02:14 PM
pplapeu

some UGLY

I earlier suggested the UGLY solution I found. Now I have posted the pics here at XS. The UGLY part really is the dual heater cores. You can see it in my link below.

As an engineer you understand the water in a water cooling solution is the heatsink.

In my solution the heatsink weighs 16.5# as it is two gallons. If you run an in case 1qt water heat sink you must rely totally on excellent rad heat exchange. I added two cores, shrouds, and 120v fans I can flip on with a lamp cord/switch. I used hardware store 4" poly 45s & tees [for two gallon res that is a dog leg with a riser to fit the round flipping coffee table enclosure-heated water enters the riser] and reduced/bushed down to the size of the cooling system, in this case 3/8". I use an I/R pump that is equivalent to an Iwaki rxl20.

http://www.xtremesystems.org/forums/...d.php?t=138100
03-22-2007, 02:22 PM
ranker

Quote:

Originally Posted by Whokidmo

Question, does air cooling the mem chips on the VGA and watercooling the main chip decrease your max OC, i mean the mem chips might over heat etc, hey OP can you do some major OCing? i honestly think air cooling and WC don't mix and that even if u do push the 2 set ups to the same speeds, 700/1200, you'll eventually burn out the swiftech card cause the mem chips get so hot with you OC, thats just my 2 bits at the end of the day, WCing is better then ACing which is why were all upgrading so why use a :banana::banana::banana::banana:ty hybird? sure the main chip temp is down but i honestly i'm like 99.9999% sure that the temp on the mem chips will increase by alot with the swiftech block or the mem chips will burn out, SHOW SOME OC TESTS.

btw i remind people many said that southbridges and vregs on mobos didn't need to by WCed... and now many disagree.

Vregs appear to be more about bling. Before settling on a 3.6O GHz OC on my e6600/evga 680i combo, I was running a 3.8 Ghz OC for a while. Vregs never really caused me any concern.

The only SB that's worth WC'ing is that of the 680i. Most other mb's don't have SB's that generate as much heat.
03-22-2007, 02:44 PM
serialk11r

Quote:

Originally Posted by ranker

Vregs appear to be more about bling. Before settling on a 3.6O GHz OC on my e6600/evga 680i combo, I was running a 3.8 Ghz OC for a while. Vregs never really caused me any concern.

The only SB that's worth WC'ing is that of the 680i. Most other mb's don't have SB's that generate as much heat.

Even that SB might not be worth it. A zalman thingy that costs 5 bucks will do the job fine (I mean its an SB, do you really need awesome cooling?).
03-22-2007, 03:14 PM
Mallen

Quote:

Originally Posted by serialk11r

Even that SB might not be worth it. A zalman thingy that costs 5 bucks will do the job fine (I mean its an SB, do you really need awesome cooling?).

The southbridge on the 680i and 650i chipsets do need good cooling to get good overclocks. I am running a p5n32-e sli plus and with my water cooling down at the moment I need to run a box fan at the side of the case to keep the southbridge cool enough to hold stable.
03-23-2007, 03:28 PM
Whokidmo

Quote:

Originally Posted by Mallen

The southbridge on the 680i and 650i chipsets do need good cooling to get good overclocks. I am running a p5n32-e sli plus and with my water cooling down at the moment I need to run a box fan at the side of the case to keep the southbridge cool enough to hold stable.

agreed so don't you think soon if not now the mem chips on VGA boards will need to be WCed for good OCs?