Bolt-on RAM/VRM sinks for 4850/4870?

[dnottis]

Using the small sinks from the 8800GTX and a low speed yate...




Cats 8.7 - core @ 850


Cats 8.8 - core @ 850


My temps are about the same as iandh with his sinks on there. So back to where we were before, it's the drivers and Furmark. The sinks dont have to be huge.

[iandh]

My temps are about the same as iandh with his sinks on there. So back to where we were before, it's the drivers and Furmark. The sinks dont have to be huge.

Thanks for that, nice to have a confirmation of my results.

Yeah, I'm using ceramique as well. As for air flow, I'm using a MM U2-UFO case with a Thermochill PA120.3 with 3 120mm Yate Loon D12SM-12 pushing air through the rad & 2 pulling air on the middle and bottom of the rad. The fan in the middle of the rad allows air to flow over the top of the card, while the bottom fans allows air to flow over the bottom (where the sinks are).

BTW, is there a program to monitor the temp of the memory on the card? I want to make sure I have good contact all around the ram. When I was putting the sink on the ram, I had a good look to see if light was getting through between the sinks and ram. So I believe have a made good contact, but I just want to make sure.

Cool, your results are similar to mine. In my WC'ed case I had air blowing on the card from a filtered front intake. I usually saw VRM loads around 90C. With the Accelero and the low speed Yate blowing directly on the sinks, I was getting about 10C cooler, as posted in the GPU-Z furmark screenshot.

So basically they should have about as much OC'ing headroom as the airflow you provide.


The ramsinks are tricky, there were only three usable mounting holes so you have to tighten them very lightly and carefully. Once you have them on right they are fine. Lots of TIM doesn't hurt either.

[ryan92084]

so i think the only question left is when an i buy these from petras
*edit* never mind petras has them they just aren't in the gpu section like the rest of iandh's work is ... odd

[dnottis]

Just as a follow up...

Still running Cats 8.8 with it setup like this -



And a low speed yate sitting on it. VRM temps hit ~80 - 84c during Furmark. Down from the 109c with Cats 8.7.



I removed the fan sitting over the top of the card and just had the 2 x 120mm blowing over the fan as seen here. (in this pic the fan is still on the HD4870, but I'm trying to show the location of the other 120mm that blow air across my PC).

So without the fan on the card, but the two other ones not so far away Furmark went to 117+C. This proves that airflow in a case is not enough to cool the VRMs. You will need specific airflow OVER the VRMs. Now keep in mind my card is running 850 Mhz core. If you are overclocked you need more airflow than just the case flow.

[iandh]

I removed the fan sitting over the top of the card and just had the 2 x 120mm blowing over the fan as seen here. (in this pic the fan is still on the HD4870, but I'm trying to show the location of the other 120mm that blow air across my PC).

So without the fan on the card, but the two other ones not so far away Furmark went to 117+C. This proves that airflow in a case is not enough to cool the VRMs. You will need specific airflow OVER the VRMs. Now keep in mind my card is running 850 Mhz core. If you are overclocked you need more airflow than just the case flow.

Those temps disagree with my watercooled card, which was around 90C with two fans mounted in the same style, but twice as far away as yours and undervolted to 1000RPM.

If I had to guess, I would say it is because the fin density is so high on the sinks that you have mounted. That high fin density only works well with direct cooling/high airflow situations. For any effectiveness in low/medium airflow situations, you need a lower fin density like on my sinks. With high airflow, I would say my sinks and the ones you have mounted would perform much more similarly.

Do you have any way of measuring your internal case temperature with the side panel on?




edit: Just fyi the sinks could be four times larger, and probably would still get over 100C without the direct cooling. Just the name of the game with this card. Anyone wanting to agressively OC should be running monster sinks or FC blocks.

[dnottis]

Those temps disagree with my watercooled card, which was around 90C with two fans mounted in the same style, but twice as far away as yours and undervolted to 1000RPM.

If I had to guess, I would say it is because the fin density is so high on the sinks that you have mounted. That high fin density only works well with direct cooling/high airflow situations. For any effectiveness in low/medium airflow situations, you need a lower fin density like on my sinks. With high airflow, I would say my sinks and the ones you have mounted would perform much more similarly.

Do you have any way of measuring your internal case temperature with the side panel on?

edit: Just fyi the sinks could be four times larger, and probably would still get over 100C without the direct cooling. Just the name of the game with this card. Anyone wanting to agressively OC should be running monster sinks or FC blocks.

keep in mind mine is at 850 Mhz core though. And I do need a bigger sink however, since I have the ram cooled already I dunno if its worth $40 just for the VRM sink. Might have to mod an old cooler I've got and cut a nice chunk to cool my vrms. BTW those are low speed yates at 100% blowing at the card. The inside of the case never gets warm, I've got alot of fans moving air out the back.

[dnottis]

Here is 900 core, no vmods, just flashed with Asus TOP bios.

[iandh]

Cool, thanks again for posting all of these numbers.

[evilrad]

Finally finished overclocking my new system and I just got around to overclocking the video card. With the GPU at 850 and memory at 1100, the VRM temps went up to 110. I might try a 900 GPU clock to see how much higher those VRM temps go.

EDIT: VPU kicked in at 900. So I guess I'll just stay with 850.

[iandh]

Finally finished overclocking my new system and I just got around to overclocking the video card. With the GPU at 850 and memory at 1100, the VRM temps went up to 110. I might try a 900 GPU clock to see how much higher those VRM temps go.

EDIT: VPU kicked in at 900. So I guess I'll just stay with 850.

For fun could you stick a fan right next to the card just to see how it affects the temps?

[Justintoxicated]

While these look great for air cooling the card, I might have to opt to replace my new DD Maze5 with a full cover block if I decide to upgrade my 8800GT 512. I can create enough airflow for a single card, but if I go crossfire the airflow from one card will be blocked quite a bit. I'm just not sure it's worth taking the risks, while I'm also not stoked about dumping VRm heat into my loop. I have your sinks on my 8800GT card, and have not had any problems but I can can nearly burn my fingers when touching them. I probably don't have optimal airflow under my video card and there is no way to attain any when running SLI short of rigging up fans hanging from the card, which defeats some of the purpose I am watercooling it to begin with. Damn, either get hot video cards, or get stuck with Nvidia Chipset.. I should have waited for next gen stuff..

[evilrad]

For fun could you stick a fan right next to the card just to see how it affects the temps?

I was going to test that out, but I actually forgot about doing it until your post. I'm going to bed now, so I'll test it out tomorrow and post my results.

[iandh]

While these look great for air cooling the card, I might have to opt to replace my new DD Maze5 with a full cover block if I decide to upgrade my 8800GT 512. I can create enough airflow for a single card, but if I go crossfire the airflow from one card will be blocked quite a bit. I'm just not sure it's worth taking the risks, while I'm also not stoked about dumping VRm heat into my loop. I have your sinks on my 8800GT card, and have not had any problems but I can can nearly burn my fingers when touching them. I probably don't have optimal airflow under my video card and there is no way to attain any when running SLI short of rigging up fans hanging from the card, which defeats some of the purpose I am watercooling it to begin with. Damn, either get hot video cards, or get stuck with Nvidia Chipset.. I should have waited for next gen stuff..

Yeah, I have to say that anyone with limited airflow or wanting to be doing heavy OC'ing should opt for an FC block.

As I had mentioned above, it doesn't have to do with the size of the sinks, it has to do with the airflow removing the heat. Even giant sinks could get screaming hot in a case with low airflow, so in that case unfortunately you are better dumping the heat in your loop.

The 48xx's are amazing cards but they sure get spicy hot when you crank them up.

[dopestuff]

hehe i don't know why everybody is talking so big about the radeon 4870 being such a HOT card....
i received mine today and ran some benchmarks and played some games...
howevver the gpu is always @ the high 60 low 70 degrees (centigrade)
and the vrm's are always somewhere in the 50 sometimes hitting 60 degrees (centigrade)
haven't tried overclocking it yet but the temps seem pretty fine by me (i was expecting worse)
and my amb is just over 20 degrees and the fan of the card is prettymuch the whole time inoudible obove my med (1400 rpm) fans
was thinking of wcing this baby but now i am going to wait a litle while longer

(running catalist 8.8)

greats, dope

[ShadedNine]

hehe i don't know why everybody is talking so big about the radeon 4870 being such a HOT card....
i received mine today and ran some benchmarks and played some games...
howevver the gpu is always @ the high 60 low 70 degrees (centigrade)
and the vrm's are always somewhere in the 50 sometimes hitting 60 degrees (centigrade)
haven't tried overclocking it yet but the temps seem pretty fine by me (i was expecting worse)
and my amb is just over 20 degrees and the fan of the card is prettymuch the whole time inoudible obove my med (1400 rpm) fans
was thinking of wcing this baby but now i am going to wait a litle while longer

(running catalist 8.8)

greats, dope

Because the type of people you find here aren't satisfied if it runs cool playing games available right now. Instead, the goal is to push every component possible with a battery of tests and then check the results. In this case, the test in question that causes these temps is Furmark, and the components being pushed over the edge are the VRMs. Quoting temps from "some games" isn't really saying anything more than quoting your idle temps.

With stock cooling, my 4870's VRMs were getting to just under 100C, and I didn't run it all that long in the interest on not playing with fire. GPU was never a problem.

[Sparky]

OK long time I know, but school mail sucks and held my package from Petra's forever, then I had no water... Anyways, got my 4870 under water with iandh sinks. I put a small low-RPM 80mm fan in the PCI slots blowing into the case. Cranked up furmark 1440x900 with other settings default. VRMs got to 98C and I stopped furmark. I think I need more airflow than this slow 80mm fan is giving me, or at least make it a bit more direct. VRMs isn't getting much. Idle is 52C on the VRM. GPU core temp is icy cold.

By the way, that vitec chip gets pretty darn hot, but has no cooling on it. Only cooling it got with stock is it kinda had maybe a tad of airflow from a slit over the stock fan... I don't know how hot it is getting but it is quite uncomfortable to the finger. I think I'm going to stick a couple of my leftover enzotech sinks on it.

[iandh]

OK long time I know, but school mail sucks and held my package from Petra's forever, then I had no water... Anyways, got my 4870 under water with iandh sinks. I put a small low-RPM 80mm fan in the PCI slots blowing into the case. Cranked up furmark 1440x900 with other settings default. VRMs got to 98C and I stopped furmark. I think I need more airflow than this slow 80mm fan is giving me, or at least make it a bit more direct. VRMs isn't getting much. Idle is 52C on the VRM. GPU core temp is icy cold.

By the way, that vitec chip gets pretty darn hot, but has no cooling on it. Only cooling it got with stock is it kinda had maybe a tad of airflow from a slit over the stock fan... I don't know how hot it is getting but it is quite uncomfortable to the finger. I think I'm going to stick a couple of my leftover enzotech sinks on it.

Haha, think how hot that Vitec chip would get under the stock cooler... it is basically sitting in an oven. Good call on stopping at 98C, any hotter is danger zone with these cards.

I recommend at least a 40CFM fan blowing unobstructed over the card from a maximum of 8-10 inches away. For most cases with 120mm front intakes or side panel fans this isn't particularly unusual. This isn't totally for my sink's benefit though, the entire card runs hot in general and all of the components, vitecs included can use a nice breeze.

Even though this is a redundant statement with all of my posts so far in this thread, it is about moving the heat away from the card, not the size of the sinks. That said, even with poor airflow these should be fine for gaming as even the 8.8's run hotter on furmark than in normal situations.

[Gir92]

I assume you are using the 8.8 driver yes Sparky?

[Sparky]

Well for now I stuck a 120mm yate medium fan turned down to around 1200RPM propped on my tubing right next to the card Also put leftover enzos on the vitec chip and the other two chips next to it. Idle VRM is now a measly 41C, load is having a hard time breaking 80C.

My personal opinion: taller sink would probably help keep cooler, but I'm sure would add to the cost quite a bit. Overall I'm happy, once I get a more permanent fan setup things will be just fine. I may get an enzotech mosfet cooler at some point or something like that and mod it to fit the 4870, to compare how much different it would be in temps.

*edit* yes I'm using 8.8.

[dnottis]

Well for now I stuck a 120mm yate medium fan turned down to around 1200RPM propped on my tubing right next to the card Also put leftover enzos on the vitec chip and the other two chips next to it. Idle VRM is now a measly 41C, load is having a hard time breaking 80C.

My personal opinion: taller sink would probably help keep cooler, but I'm sure would add to the cost quite a bit. Overall I'm happy, once I get a more permanent fan setup things will be just fine. I may get an enzotech mosfet cooler at some point or something like that and mod it to fit the 4870, to compare how much different it would be in temps.

*edit* yes I'm using 8.8.

I had a similar setup. After 3 weeks the card died. A 120 was sitting right over the card, VRM, core temps all in check too. Good luck with yours.

I feel this is very important and needs to be shared. While dealing with the HD4870 I was able to pose some questions to an engineer at ATI. One of the responses that concerned me most was this one -

“when a user replaces our stock cooler with something that is more capable (eg water cooling or higher capacity integrated fan), RV770 XT GPU will draw significantly more current and cause the regulator to hit its temperature limit at 125C.”

Now basically what he is saying was that when not using the stock fan, the card will draw more current that it otherwise would. Alot of people have blamed the deaths of the HD4870s on me water cooling them, I guess it's possible and imo this may have been the reason. The part of the VRM hitting 125C was pre 8.8 Catalyst where the VRM would almost shoot up to 125c right away. I believe the important part to take away from this is that the HD4870 without the stock fan connected to the header can pull more current than it would with the 3 pin header connected to the PCB, possibly the reason my two cards died. Be careful.

[iandh]

I had a similar setup. After 3 weeks the card died. A 120 was sitting right over the card, VRM, core temps all in check too. Good luck with yours.

I feel this is very important and needs to be shared. While dealing with the HD4870 I was able to pose some questions to an engineer at ATI. One of the responses that concerned me most was this one -

“when a user replaces our stock cooler with something that is more capable (eg water cooling or higher capacity integrated fan), RV770 XT GPU will draw significantly more current and cause the regulator to hit its temperature limit at 125C.”

Now basically what he is saying was that when not using the stock fan, the card will draw more current that it otherwise would. Alot of people have blamed the deaths of the HD4870s on me water cooling them, I guess it's possible and imo this may have been the reason. The part of the VRM hitting 125C was pre 8.8 Catalyst where the VRM would almost shoot up to 125c right away. I believe the important part to take away from this is that the HD4870 without the stock fan connected to the header can pull more current than it would with the 3 pin header connected to the PCB, possibly the reason my two cards died. Be careful.

I'm not in any way trying to be rude here, but until I see some type of white papers, what that engineer said sounds like complete and utter crap.

I've worked at an electronics manufacturer for the last ten years. I've taken multiple classes in AC and DC electrical theory, and solid state circuit design. I haven't cared to finish a degree yet, but I'm not that far, trust me.

There is no logical way that a card would "sense" that it had its fan removed, and then for some reason all of a sudden start sucking more power. It just doesn't add up to me. There is no way that I know of that cooling a component will suddenly cause it to start drawing more current, nor can I see any reason that they would design the card to self destruct with aftermarket cooling which is basically what you have quoted from the AMD engineer.

What I am guessing is that this guy does not quite know what he is talking about. Perhaps he means that when people put aftermarket cooling on and overclock the card it can still overheat, but even that doesn't make any sense from what I've seen.


Remember, I've had two (now three) cards testing for going on two months now running perfectly cool and without the slightest sign of stress or strain. The reported current draw in gpu-z never increased upon removal of the stock cooler on any of the cards.

[dnottis]

I'm not in any way trying to be rude here, but until I see some type of white papers, what that engineer said sounds like complete and utter crap.

I've worked at an electronics manufacturer for the last ten years. I've taken multiple classes in AC and DC electrical theory, and solid state circuit design. I haven't cared to finish a degree yet, but I'm not that far, trust me.

There is no logical way that a card would "sense" that it had its fan removed, and then for some reason all of a sudden start sucking more power. It just doesn't add up to me. There is no way that I know of that cooling a component will suddenly cause it to start drawing more current, nor can I see any reason that they would design the card to self destruct with aftermarket cooling which is basically what you have quoted from the AMD engineer.

What I am guessing is that this guy does not quite know what he is talking about. Perhaps he means that when people put aftermarket cooling on and overclock the card it can still overheat, but even that doesn't make any sense from what I've seen.


Remember, I've had two (now three) cards testing for going on two months now running perfectly cool and without the slightest sign of stress or strain. The reported current draw in gpu-z never increased upon removal of the stock cooler on any of the cards.

I know where you are coming from, cause my reply to that email went something like this "if that is true, that is the dumbest thing you guys could have done. With the amount of heat being produced do you really think people aren't going to use third party or water cooling on these???"

They didn't believe most people would change the cooling. The only way I could conceive the card would be able to "sense" that the stock cooler was changed was through the 3-pin fan header. Obviously there is nothing else that would tell the card the cooler was not installed. Alot of emails went back and forth discussing which parts needed cooling and what the failure temperatures were, I thought that information was something that I should share now that I'm not even using an HD4870. That was a direct paste from the email I received though. I hope there isn't a bigger issue and these cards don't all start failing. Maybe I just got really unlucky with the two I had.

Anyways, I'm still hoping you put out a set of sinks for the GT200 series iandh I've had to ghetto chop the 8800GTX mosfet sinks for my GTX 260.

[Sparky]

As an EET major that sounds mixed up to me as well.

[iandh]

I know where you are coming from, cause my reply to that email went something like this "if that is true, that is the dumbest thing you guys could have done. With the amount of heat being produced do you really think people aren't going to use third party or water cooling on these???"

They didn't believe most people would change the cooling. The only way I could conceive the card would be able to "sense" that the stock cooler was changed was through the 3-pin fan header. Obviously there is nothing else that would tell the card the cooler was not installed. Alot of emails went back and forth discussing which parts needed cooling and what the failure temperatures were, I thought that information was something that I should share now that I'm not even using an HD4870. That was a direct paste from the email I received though. I hope there isn't a bigger issue and these cards don't all start failing. Maybe I just got really unlucky with the two I had.

Anyways, I'm still hoping you put out a set of sinks for the GT200 series iandh I've had to ghetto chop the 8800GTX mosfet sinks for my GTX 260.

Don't get the idea that I don't appreciate your contribution, keep it coming...


It just sounds wierd. Is there any chance that english is not his primary language and there is some type of misunderstanding?


I talked to my boss about switching voltage regulators, and he said that it is normal to see a current spike before they fail. Other than that, the current draw should stay dead stable regardless of temp up to the failure point. If the VRM's were left uncooled with an aftermarket cooling solution for the GPU, it would be expected to see a spike in current draw followed by failure, as we've seen several times here.

Maybe that is what he is getting at.

As an EET major that sounds mixed up to me as well.

Yeah, there has to be something here somebody is missing, I'm not sure who or what though.

[Aldy402]

iandh

i have your 8800gt sinks and they rock, they allow longevity for my mcw60 block. 1 thing though..

could you machine us a 120/140mm fan bracket that utilizes the pci cover bracket slots? It would also be cool if it was adjustable from side to side...

that would really help a lot of us using just sinks or your sinks in conjunction with a waterblock