Asus HD4850 new cooler issues

[pavithra_uk]

fixed it.
reason : one screw in Cooler was get loosed and GPU & heatsink contact become loose.
fasten it. all fixed. now temp low as 70C max load.

 

[silkstone]

Your card does seem a little hot, although i have heard of 4850's getting hotter. I wouldn;t worry about the Ram or VRM sinks unless you are overclocking, and i don;t think the 4850 can be Vmodded to increase the voltage anyway, so the VRMs should be fine.

The disparity in the 3 temps you shown is usually due to incorrectly mounted heatink. i'd remove it, clean it, apply some decent TIM and re-seat it.

 

[Widjaja]

Yeah that cooling system sucks.

I have always been baffled by ASUS choice to remove the stock cooling and place just GPU cooling.
Then ASUS released the Rev 2 of the 4850 which has terrible customer reviews if you checkout newegg.

Your temperatures are as stated a few degrees cooler but this is on the GPU core side.
You can not read the temp of the VRMs.
I personally think the VRMS are indeed running hotter than with a stock cooler.
4850s which have overheating VRMs will cause black screen and VPU recover.

I personally never trusted the ASUS cooling on these cards.

You cold always go for a Arctic Cooling Twin Turbo or fabricate up your own aluminium heatsink for the VRMs.

Just hack/grind down some old CPU or Chipset heat sink, and get some thermal tape.
Don't really need to worry about the inductors which are the grey squares

I wouldn't be too worried about the RAM was the GPU heatsink is designed to blow air on to the RAM.

 

[pavithra_uk]

..

 

[Widjaja]

There is air blown on to the RAM chips which is whats important.
It may not be optimal.
You will see the fins of the cooler are set out in a way for the air to blow on to the rest of the card.
If you were to use RAM sink cooling modules, you will most likely need to replace the cooler as well.

I do not know of any RAM cooling modules which are able to fit under the cooler.
Zalman and Thermal Right do have low profile RAM cooling modules but you maybe paying more than you would want.

 

[pavithra_uk]

..

 

[Widjaja]

I overclocked Memory to 1050 MHz (stock 990 MHz) then run ATI tool for artifacts. Artifacts appear after 10 seconds.

If such small increament (60MHz) can Artifacting may indicate instability of VRAMs ?

My HD4850 is the same and I have memory cooling modules on mine.
Mine is fine at 1050Mhz but any further and it artifacts yet I can overclock the core up to 700Mhz no problem.

Sure some state they can run at 1100Mhz or higher though.

Stock speeds & Furmark run for 10 minutes after Display sometimes go black. ---> VRM overheat ?? (No VRM heat sinks on this card)

I already attached a Heat sink for VRM. Grab heat sink from a NB cooler & thermal pads from few old CD-RW,DVD-ROM drivers.

What?
You told me you had cooled your VRMs.

If I RAMed the card, Asus will send me same card & new faults as a bonus...

same thing happens to card before (black screen @ games) due to Reseller mistakes, it took 5 months to RMA process. so reseller give me 2 months extended warranty.

Already ask about this matter in ASUS Forum. not even single reply.

Install an accelero S1 or a Accelero twin turbo.
Or sell it for a different card.

 

[silkstone]

Vrm's would have nothing to do with artificating at high memory speeds. In my experience it's just the quality of the chips, they really don't overclock very well on 4850's. I have killed one before by cranking the memory up too high.

I wouldn;t worry too much about overclocking the memory, the biggest benifits come from the core.

oh, and did you manage to get the card any cooler than what u stated in your op?

edit - Don;t put any heatsinks on the vrms if u don't know what you are doing, you wouldn;t want to short them and end up with a dead card.

 

[DarkOCean]

That pcb seems reference and i wouldn't dare to let the vrms without a heatsink on them. Memory oc depends on my 4850 depends on the core voltage at 1158 mv i coud hit 1100 no problems ,at 1121mv 1050 mhz and at stock 1084 mv 1010 max.

 

[Widjaja]

That pcb seems reference and i wouldn't dare to let the vrms without a heatsink on them. Memory oc depends on my 4850 depends on the core voltage at 1158 mv i coud hit 1100 no problems ,at 1121mv 1050 mhz and at stock 1084 mv 1010 max.

Who knows what ASUS were thinking when they decided it would be fine to run these cards without VRM cooling.
They also released a revised PCB version which was even worse than the last according to newegg customer reviews.

I believe MSI also released a HD4850 without VRM cooling as well.

I have ran my HD4850 with passive cooling for a test and at 85degC it black screened which I believe is the VRMs overheating which is most likely true as one of my VRM coolers was elevated by a resistor at the time causing one of the VRMs to have no heat dispersion.
AC did not mention this would be an issue, just like having to bend the fins on one of the cooling modules.

 

[micropage7]

Vrm's would have nothing to do with artificating at high memory speeds. In my experience it's just the quality of the chips, they really don't overclock very well on 4850's. I have killed one before by cranking the memory up too high.

agree
but what about your case airflow? or ambient temperature
over 70 degrees is kinda pretty worrying

 

[Widjaja]

agree
but what about your case airflow? or ambient temperature
over 70 degrees is kinda pretty worrying

70degC is pretty cool for a HD4850 and about right for the cooling provided on the OP card.

I think I may have covered this in a previous post.
My reference model HD4850 idled at 84degC and hit 110degC at max load but still ran like a champ but I was concerned for how long so I installed an AC Twin Turbo on it.

The reason the card idled so high was due to the autofan setting in the GPU BIOS.
The card ran passive until 70degC.

It's just the concern about the rest of his card since there is no cooling modules on the VRAM and VRM besides what air the GPU cooler provides.

 

[micropage7]

70degC is pretty cool for a HD4850 and about right for the cooling provided on the OP card.

I think I may have covered this in a previous post.
My reference model HD4850 idled at 84degC and hit 110degC at max load but still ran like a champ but I was concerned for how long so I installed an AC Twin Turbo on it.

The reason the card idled so high was due to the autofan setting in the GPU BIOS.
The card ran passive until 70degC.

It's just the concern about the rest of his card since there is no cooling modules on the VRAM and VRM besides what air the GPU cooler provides.

110 degrees? you can cook there
thats why i dont like auto setting. you cant check every single thing just choose and leave it. but if you dont wanna sweat its the solution

 

[silkstone]

That pcb seems reference and i wouldn't dare to let the vrms without a heatsink on them. Memory oc depends on my 4850 depends on the core voltage at 1158 mv i coud hit 1100 no problems ,at 1121mv 1050 mhz and at stock 1084 mv 1010 max.

Who knows what ASUS were thinking when they decided it would be fine to run these cards without VRM cooling.
They also released a revised PCB version which was even worse than the last according to newegg customer reviews.

I believe MSI also released a HD4850 without VRM cooling as well.

I have ran my HD4850 with passive cooling for a test and at 85degC it black screened which I believe is the VRMs overheating which is most likely true as one of my VRM coolers was elevated by a resistor at the time causing one of the VRMs to have no heat dispersion.
AC did not mention this would be an issue, just like having to bend the fins on one of the cooling modules.

I have a gigabyte 4850 lying around, it came with a zalman fan but no VRM HS. It also supports overvolting upto 1.20v or something. i never had a problem with overclocking the core, and the hottest it would get would be 70-80 depending on what i had my fan speed set to. i would run the core at 750 24/7 no problem, but the ram on 4850's is shite, i would only dare to clock it 50-100mhz over default.

The "standard" operating temp on reference 4850's is a bit around 80-100C btw. I remember doing some research and was surprised at how cool mine ran even when overclocked.

 

[pavithra_uk]

..

 

[Widjaja]

110 degrees? you can cook there
thats why i dont like auto setting. you cant check every single thing just choose and leave it. but if you dont wanna sweat its the solution

Yeah it wasn't fun.
The back corner of my case was near on impossible to touch after a short while.
There was one game which always brought it to 110degC.
The cartoon Star Wars Knights of the republic at max settings
The graphics were basic but it stressed my card none the less.
My solution was to upgrade the cooler, edit and flash the GPU BIOS so the twin turbo fans would start at 60% fan speed.
Otherwise they would not kick on at all without software.

I have a gigabyte 4850 lying around, it came with a zalman fan but no VRM HS. It also supports overvolting upto 1.20v or something. i never had a problem with overclocking the core, and the hottest it would get would be 70-80 depending on what i had my fan speed set to. i would run the core at 750 24/7 no problem, but the ram on 4850's is shite, i would only dare to clock it 50-100mhz over default.

The "standard" operating temp on reference 4850's is a bit around 80-100C btw. I remember doing some research and was surprised at how cool mine ran even when overclocked.

The card you have is non reference PCB.
Gigabyte may have done a nice job on the power phase revision of the card over the reference which kept the VRMs cool enough not to need a VRM cooler.

In my case the memchips were indeed crap.
One name synonymous with poo chips......Quimonda.

When I removed the stock cooler and saw them i knew there was a likely hood of getting a poor mem OC.

Even so there are people who have apparently ran stable at 1100Mhz with the card I have.

I increased the voltage on mine at one stage and it did squat.
Still artifacts.
Oh well at least I don't fidgit around in my seat due to the heat my card was dishing out.

yes. I removed DIY heatsink that I installed. and run stability test.

the heatsink I installed is that come for NorthBridge cooling.
Furmark with heatsink and 10 minutes runing time never experienced black screens.

this heat sink getting damn hot. can't even touch.
same amount of heat produce in RAM chips after furmark or ATI tool artifact test.


I think asus failed..
such heat never kill RAM chips or VRM MOSFETS instantly, but it greately reduce componant life expectancy.

ASUS must add following sentence to their VGA box:
"USER MUST INSTALL 3RD PARTY VRM & RAM HEAT SINK IF THEY NEED STABLE CARD"

I would expect the heatsink to get hot but that is showing there is heat being transferred and dispersed which is much better than none in your case.
My stock cooler had warning stickers and one of them was right next to the VRM cooler.

I would look for a decent heatsink and use sekisui thermal tape to attach it to the VRMs.
An accelero S1 and add a fan to it or a twin turbo would be your best bet but I don't think Arctic Cooling makes the standard Twin Turbo anymore.
Only the Twin Turbo Pro with 92mm fans.
You will also find with the Twin Turbo Pro, the copper pipes run slightly above a couple of VRAM chips which leaves no space for adding even low profile cooling modules to.
You will most likely have to edit the GPU BIOS as well to get the fans to run if you do not want to rely on software for your fans to run.

 

[silkstone]

Yeah it wasn't fun.
The back corner of my case was near on impossible to touch after a short while.
There was one game which always brought it to 110degC.
The cartoon Star Wars Knights of the republic at max settings
The graphics were basic but it stressed my card none the less.
My solution was to upgrade the cooler, edit and flash the GPU BIOS so the twin turbo fans would start at 60% fan speed.
Otherwise they would not kick on at all without software.



The card you have is non reference PCB.
Gigabyte may have done a nice job on the power phase revision of the card over the reference which kept the VRMs cool enough not to need a VRM cooler.

In my case the memchips were indeed crap.
One name synonymous with poo chips......Quimonda.

When I removed the stock cooler and saw them i knew there was a likely hood of getting a poor mem OC.

Even so there are people who have apparently ran stable at 1100Mhz with the card I have.

I increased the voltage on mine at one stage and it did squat.
Still artifacts.
Oh well at least I don't fidgit around in my seat due to the heat my card was dishing out.



I would expect the heatsink to get hot but that is showing there is heat being transferred and dispersed which is much better than none in your case.
My stock cooler had warning stickers and one of them was right next to the VRM cooler.

I would look for a decent heatsink and use sekisui thermal tape to attach it to the VRMs.
An accelero S1 and add a fan to it or a twin turbo would be your best bet but I don't think Arctic Cooling makes the standard Twin Turbo anymore.
Only the Twin Turbo Pro with 92mm fans.
You will also find with the Twin Turbo Pro, the copper pipes run slightly above a couple of VRAM chips which leaves no space for adding even low profile cooling modules to.
You will most likely have to edit the GPU BIOS as well to get the fans to run if you do not want to rely on software for your fans to run.

Yup, but if the op is inexperienced, i wouldn;t recommend cooling the VRMs. There would be no benefit to memory clocks, it might help with core clock but a bigger/better hsf would be more effective.

 

[Widjaja]

Yup, but if the op is inexperienced, i wouldn;t recommend cooling the VRMs. There would be no benefit to memory clocks, it might help with core clock but a bigger/better hsf would be more effective.

I think he would need more experience installing a complete GPU cooler than sticking on some memory and VRM cooling modules with thermal tape.

I saying that I personally think it would be the best bet to get a nwe cooler as I have said.
The AC accelero S1 and AC Twin Turbo both come with a set of cooling modules which will cover all VRAM and VRM chips with a little alteration.