Sure will - tonight I'll post results
Printable View
Ok so I removed the sink - kept the 50mm on there, Furmark ran ok. Removed the little 50mm fan grabbed a med speed yate loon, screwed on a fan grill and put this over the ass end of the card. Ran this - the VRM temps were a few deg lower than the 50mm was keeping it. Looks like those little 3 on the bottom are really the culiprits but they need alot of air. Without air the heatsinks get about 150 deg F - as per my handheld thermal gauge. They were burning my fingers and I couldnt hold it on there long. Alot of air is absolutely necessary - a little wont cut it. Also, the ram gets damn hot too during furmark.
http://img509.imageshack.us/img509/3...800x600gv3.jpg
http://img518.imageshack.us/img518/1...800x600jy7.jpg
new everest beta -
http://www.lavalys.com/beta/everestu...axy7stgf2c.zip
All this seems to say that GPU only WB won't be a good deal for such a card if you have to keep a high speed fan just dedicated for it :mad:
yea, but I had the same crashing with the stock cooler - which means a full cover block would need to actually have the water move over the VRMs.
EK doesn't even have water going over the VRMs -
http://www.xtremesystems.org/forums/...&postcount=147
Do you guys notice if the memory gets really hot on your cards during Furmark?
I've noticed in games that the VRM goes up to around 50C. While GPU only goes up a few C. I honestly wouldn't worry about Furmark as it really doesn't represent the kind of stress your video card will experience in most games.
I agree, but...
It's like asking someone to use a processor crashing after 10mn of prime95 because quiet no program will stress it that much :confused:
For me, it's hard to get a product that won't be able to work when it is asked to do at 100%
But surely, for many people outside XS and overclocking forums, it won't be a problem since they never heard of Für. That's maybe 90% of PC users :D
My VRM idles aboout 50c with huge heatsinks and airflow. RTHDRIBL load it to about 80c. Games its about 70c. Then again we dont even know how accurate the Everest sensor reading is.
Im with Johnny about stability - if its capable of crashing then thats more than enough for me to do something about it and address it. I can't just say oh well it's just Furmark.
Another good reason to just get a GTX260. Though I don't support one side or the other. I think they're both good, both have their pros, both have their cons.
It was my understanding that the problem dnottis was having was vrm heat related with just Furmark.
As long as he is able to keep it under control he should not have any problems. GPU-Z or Everest revealing temps offered more insight to the problem then Furmark. If your games are running fine then Furmark offers no tangible benefit. This is my take on the situation so far. If I missed something then my apology.Quote:
Here's a pic of mine with the MP-01. Idle 32, load 38c. The only issue I have is Furmark will crash, always had even stock settings / cooling. Would get a black screen after about a minute. I can game for hours without an issue.
This is where technology hurts. Make it smaller and smaller till you get a chip so small its hard to cool. Kinda like 45nm vs 65nm - my wolfdales all loaded higher than my 65nm quads. They keep shrinking everything to the point where there is a spot the size of the tip of a pen that you need to cool. If the VRMs had some more surface area they'd be easier to cool...
I had to make the VRM sinks larger on this model compared to the 8800GT/GTS/9800GTX versions of my sinks.
I have been running my Visiontek on the stock heatspreader for probably two weeks along with an MCW60, with the stock blower removed (so no additional airflow other than my case's), and not a single hiccup. Although the board as a whole runs toasty, all the components seem to be reasonably uniform in temperature. Nothing I would call out of hand.
Perhaps I have a cooler running card, also with issues like this it is very hard to draw the line between testing and intended use. As mentioned before, a particular stress test may push a card to the bleeding edge and over it. It is possible that this card is only certified to something like 80% power draw @ full duty cycle... also hard for me to say how furmark compares to the testing methods that ATI uses to call a card "good".
Sadly my HD4870 died tonight. I got some corruption while playing MOHA, then back at the desktop it just went green with some lines. RIP.
will this cover all critical areas
http://ekwaterblocks.com/shop/produc...roducts_id=289
Oh, sorry for you. That was the second card or the one you wanted to RMA?
So heat could kill those cards. Any one used to electronics knows that even if some circuits are built to run at 125°C, they will for sure die quicker than if ran at 60°C or 55°C
One more reasone not to cut original heatsink, or your RMA will be voided
Maybe also mounting/unmounting without electrostatic protection killed your card :rolleyes:
That was the first one. So maybe the VRM overheating issue is a sign of faulty components on the GPU. I didnt remount it before it died. It just started to artifact in MOHA, when I came back to the desktop within a few minutes the screen went green with lines. Would not boot anymore, no more LEDs, nothing. Dead GPU. I had the fans over the back end and the VRMs werent getting all that hot when it died.
Definitely dont cut up the stock heatsink. Heres hoping an Asus 4870 will be a better choice.
Wow...a LOT of info in this thread.
I am in the planning stages of a watercooling setup in my Shuttle XPC SP35P2 that won't probably happen for a couple months, but I was looking at GPU blocks and was considering the D-Tek Fuzion v2 GPU block when someone pointed me to this thread.
I suppose it's a good thing that it'll be a little while before I'm ready to start buying/building this loop...hopefully things will be sorted out by then.
I'll be keeping a close eye on this thread though, keep at the research guys...all the hard work is very much appreciated. :up:
That was an intense thread, and the hero died at the end too.
I'm really hoping that all this turns out NOT to be purely temperature related, as I would really like to be able to get away with a GPU only block...but I wouldnt have a lot of airflow over the card to take care of the memory/vrm and other heatsinks on the card.
Decided to get another brand after the HIS bit the dust. Asus HD4870, actually ran Furmark on stock cooler. VRM runs cooler too on this board. I sinked everything as good as I could and have the 120 over it. So far it's looking good.
I believe the HIS HD4870 was just defective.
http://img239.imageshack.us/img239/8739/img0353tz1.jpg
http://img110.imageshack.us/img110/2878/img0352pc8.jpg
http://img105.imageshack.us/img105/5822/69130818rq6.jpg
http://img375.imageshack.us/img375/3710/img0355ks9.jpg
Is it stable enough for Furmark without the 120mm fan? Chances are I'm gonna cave and go with a full coverage EK block, but if I change my mind I'd be curious to see how your new card does with the heatsinks but no direct airflow.
That's look a definately better VRM temperatures. So, there are some bad batches of these cards, since on many forums this problem is reported.
I'm also interested to see how it runs without the 120mm fan, with just adequate airflow in the case
*shrugs*
Maybe I'll just deal with the extra heat and go with the EK full coverage block when it's available.
Me2...if the VRM's and other small components can run without direct airflow then I'll get a GPU only block to lower heat dumped into the loop.
Seeing as I don't have the option of direct airflow on the card in my Shuttle I'll have to go with a full coverage block if the card can't handle it with no direct airflow so I'm eager to see what happens. ;)
A Shuttle doesn't have a lot of room to breath... :(
In dnottis situation, reversed ATX should be good.
The heat isn't trapped under the PCB...
Same here in my stacker stc01 :)
I know...the best I could do is add a 80mm or 92mm to the left side panel that blows across the system towards the video card, but it would still be very weak airflow that wouldn't hit the whole card once it got to the HD4870 anyways.
Pretty sure I should just play it safe in the Shuttle and go with the full coverage EK block when I can but I'm still curious to see what's possible with the GPU only setup.
Without the fan over the card, the VRMs get smoking hot. Hot to touch and the sensor reflects that. Here is idle and full out Furmark - without additional cooling these cards are going to burn up. even with the stock cooler, they are going to die if you push Furmark too much.
http://img514.imageshack.us/img514/8...0812bm9.th.jpg
http://img291.imageshack.us/img291/6...0813ai2.th.jpg
Yes the VRM is running at 125C...but what is the chip rated for? I could have sworn I heard it said earlier in this thread that they're good up to 125C for extended use. If that's the case then wouldn't it be ok? Obviously you're not going to be running Furmark 24/7...and no game pushes the card to those limits, so if it peaks at or just below the rated temperature during Furmark then I would think it would be ok for day to day gaming would it not?
EDIT: Just did some reading and it does seem like the VRM is supposed to be able to handle 125C just fine...so then I ask...can you do everything we USED to do to determine stability before Furmark came along? 3DMark06/rthdribl etc...I'm not saying that Furmark is botching numbers, but if it's the only program that heats the VRM up that high but even then it's still within spec I don't see what the problem is for day to day use.
Isn't it also possible that Everest is reading/reporting temps incorrectly? Has someone confirmed these 120c VRM temps with a infrared thermometer or something?
Any thoughts?
From the readings on a hand held thermal sensor I've say the VRM is hitting about 70c on the surface, I dont believe 125c, 250F cause they would just melt the PCB. The heatsinks on the VRMs get very hot to touch while running Furmark.
So, yes it's possible. I just dont want to fry another card. So I'll be using additional fans for now. Use caution as I dont really think the stock cooler can do a decent job with the VRMs getting as hot as they do.
Fair enough...I'd probably be the same way if I had just lost a card as well.
Interesting to see the difference between surface temperature and the internal reading. Whether or not the VRM temp reported by Everest is even accurate I don't know.
You're great, many thanks for doing the tests.
I now fully agree with you that using a GPU only WB without active cooling will kill th eboard at those temperatures :confused:
It's really the VRM. Sad, it seems that the time of GPU only WB is done, unless we can fit some mosfets WB on the card
I received my koolance full cover block:
http://www.xtremesystems.org/forums/...16&postcount=8
lol, you make it sound like you DON'T like paying 3 or 4 times as much for more restrictive less effective GPU cooling. :ROTF:Quote:
Expensive too
Although it does look nice, I'm gonna wait on the EK full coverage block myself...seems to have a little more potential, I dunno...I think that EK stuff just generally gets a stronger thumbs up then a lot of Koolance gear to begin with, Koolance is just the only one a lot of people have as an option for a full coverage 4870 block at the moment.Quote:
I received my koolance full cover block
I wish I would have seen this thread earlier as I already pre-ordered the Fuzion block from NCIX this past monday but now I'm gonna cancel it.
Anyway, thanks for taking the time to do all these testings. Really appreciate that. But I think the stock cooler seems to do the job just fine.
HIS HD4870, fan speed manually set @ 55%, room temp ~ 27C, stock clock.
Everything is still air cooled as I'm still bench testing my Q9450 & MF2. VRM temperature seems to stabilize in the 80's. but perhaps it's too early to tell :confused: I stopped it after 7 mins.
http://img522.imageshack.us/my.php?image=furmarkex3.jpg
So, I'm assuming then that the D-Tek isn't a single slot solution?
I'm more interested in what waterblocks are single slot solutions. I know that full cover blocks are, but is the D-Tek Fusion 2 single slot?
I think you may be jumping to conclusions here. GPU-only blocks are not 'done'. You just need good air flow to your components! The stock cooler provides direct air flow to the vregs/ram, and needs to be compensated for if you remove it.
From a failure analysis POV, if the OP's original card crashed multiple times, then finally died, thats a very good indication that one of the board's components were bad to begin with. Probably would have died if he would have kept using the stock cooler.
Flashed my Asus HD4870 today. Auto tune gave me 890 core, but I backed that off as I dont want to fry this card.
Here is a Vanatage run.
http://3dxtreme.net/other/Q6600%20L7...edAsus4870.jpg
..and 3dmark06
http://3dxtreme.net/other/Q6600%20L7...870flashed.jpg
Slick...I'm gonna have to give my HD4870 a shot...never even tried pushing it.
ok, here is the VRM temp, using the Koolance FC block.
I set the rad fans on slow and silent, this is stock speed for CPU and VGA (stability testing with prime and furmark before OC).
If I got it correctly the VRM is not cooled by the waterblock, correct?
http://i104.photobucket.com/albums/m...er/vrmload.jpg
And, we need some tests on a FC block for this card. Any temps readings under für with a FC block
http://www.fudzilla.com/images/stori...d_accelero.jpg
Im wondering how the cool te VRM's here...
When you say "stock", you mean the stock cooler ?
I think it's pretty good :
http://www.xtremesystems.org/forums/...=195882&page=2
2nd last post.
Did you have a chance to OC the card and if so, what's the temperature like on the VRM with the Koolance block installed.
I'm not really a fan of full-covered block and my MCW55 has served me well for 3+yrs on 3 different cards. But I think I'm gonna bite the bullet and spring for a full block this time.
Either this one or the EK block. I'm leaning towards the Koolance block cuz it's cheaper lol.
btw can someone tell me the safe max. mem clock ?. I managed to run it @ 1100 with 2+hrs Atitool and 40+ mins into Furmark w/o problem, other than the v. high VRM temp.
Is it true that the DDR5s shouldn't be pushed beyond 1000. I really don't wanna kill the card prematurely
especially if I am going to spend 100 bones just on the full-covered block alone. TIA
just a quick question
see this water block you think this will fit the 4870 ?
http://www.aerocooler.com/shop.cart?..._id=FANTTW0150
http://www.techpowerup.com/img/07-05...I2900XT-15.JPG 2900 Xt stock cooler
http://www.techpowerup.com/reviews/D...es/cooler2.jpg 4870 stock cooler
look at the screw hole they all look like they're in the same spot.
my chopped stock cooler does about 49C idle 85C load, sorry I don't have ss for ya ;)
My previous posts are all using the koolance full block, I didn't try the card I'm running it at 750@1100 for 2 days running furmark. Why we should keep memory under 1000?
The VRM is on the lower 70s.ºC. GPU and memory are hitting 41ºC.
Should I raise the core and lower the memory? Where do you guys read about higher than 1000mhz would kill the card?
it doesnt, all cards arent guarenteed to do over 1000, depends on the bin of the ram obv, but with volt mods ram easily does 1150 (im not even Mvddc or mvddq modded and i can do 1125 artifact free)
So what's the verdict on the 4870? Can a GPU only block with ramsinks work OK or is a fullcover EK block preferred?
Probably full-cover block is preferred, but if you don't want to spend $$$ for FC block and throw it away after card upgrade, MCW-60 + handmade sinks on RAM and VRM parts do the job just fine. I have 825/4000 rock stable without voltmod (yet?). Can post pics later...
I think I'm going to wait on iandh sinks for it :D
i'm having strange stability issues with a 4850+mcw60, but only in 3dmark.
furmark and games run flawlessly
http://www.xtremesystems.org/forums/...0&postcount=32
http://www.xtremesystems.org/forums/...6&postcount=34
i suspect that there could be any chip(on the card's back?) that needs additional cooling
4850 power circuitry is different from 4870 and i think it doesn't have those chips that you've ramsinked/cooled on 4870. any clue on what component i have to cool?
one more note, with a fan blowing at full rpm (2500) to the card i have no crashes(3dmark only), but if if lower rpm(1500) it crashes
all critical components that im aware of are ramsinked: memory chips and mosfets
Watching the advance here. Got the Dtek waterblock and heatsink and going to dangle with some of the suggestions here.
Subscribed.
Very intersting stuff! I have all the bits to install two mcw60's on a pair of Crossfired 4870's. I was thinking about bending tube and kinks between waterblocks, when I decided to research a bit more. Maybe it's worth buying FC Waterblocks. Cost of Blocks v cost of New Cards. Two Cards in CF don't leave a lot of room for Air movement!
I ran my 4870 on a MCW60R with swifty ramsinks on the ram and the 3 pulse chips, also the small alu heatsinks swifty do where on the pwm's and a 120mm fan blowing over the card, for the last 3 weeks. Although the gpu core temps where very good (32 idle-42 loaded) the same could not be said for the vrm's(idle 52), which I started another thread about. I was seeing in Crysis and Jericho 102-103c, Grid,Mass Effect,GOW mid 90s,COD4,Bioshock mid to high 80s and the usual furmark 126c within 60 sec's.
I was not really comfortable with this and in the end I bit the bullet and brought a EK FC block, which I am not a fan of due to loss of value and to me high premium of price, also the talk of restriction. However I really must say how impressed I am with it, there is no noticable restriction over the mcw60 in fact cpu temps appear to have dropped by 1 or 2 degree's when loaded, however there may other factors affecting this. But the vrm,s are being cooled much more effectively idle temps are 39-40, gpu core is the same 32c, but I am seeing no more than 63c in Crysis and Jericho, high and los 50s in the other games I mentioned above, so a drop of about 40c across the board. Furmark is now not even getting above 73c where as before it would hit 126c before 75 percent of the test(see screenie)and to be honest I am seeing lower gpu temps.
http://www.r5jones.aquiss.com/Pictures/fm.jpg
I think this is one instance in which a Fc block is maybe needed, to protect your card.
sorry to go offtopic, but did u guys see what biohead did?
http://www.xtremesystems.org/Forums/...7&postcount=36
maybe some mosfet cooler from thermalright might work on 4870/4850:
http://www.thermalright.com/new_a_pa...et_cooler.html
I thought about doing something along those lines,but took the easier option as even with a 120mm fan sat ontop it had little effect.
Good info holwill, do you have any pics of the FC installed?
Definately the GPU only solutions should be banned for now until some unisinks are available
Cool thanks.
I am deciding between getting the EK fc and waiting on Thermaltake's T-Rad2... The VRMS will get airflow from the looks of it.
http://www.expreview.com/img/review/...-rad-ins11.jpg
HR-09 Type 2 has already been done in conjunction with an HR-03 GT with great success. However, mounting it is not easy. Check it out here...
http://www.clunk.org.uk/forums/user-...hd-4870-a.html
That is definitely one way of doing it.
There you go lyl excuse the crap camera and idiot behind it
:D
http://www.r5jones.aquiss.com/Pictures/P8180194.JPG
http://www.r5jones.aquiss.com/Pictures/P8180195.JPG
ati underclocking when furmark detected. whats wrong with these cards?
http://www.geeks3d.com/?p=1106
The suggestion seems to be that Furmark loads the cards well above any real-world usage scenario, leading to dangerous temps that would not likely be seen elsewhere; and so AMD downclocks the cards for safety's sake.
or, IMO a cheaper solution might be this, as it's tall enough and would probably manage to dissipate enough heat
(although it's not Cu, and would require some cutting to properly fit)
so what happenes when a game comes that does the same as furmark? we should downclock our cards to avoid damaging them?
Both waterblocks from AC and EK perform very well. I think EK won for about 3ºC in full load (need to check my notes :D)
If anyone need i can upload some pics.
I've been posting for weeks now how it looks like there are major issues with 4870 VRM cooling and/or the overall power design...some listened, some did not...now we have official confirmation...
...sadly ATI just "glazes" over this by hacking the driver and not fixing a problem...
Now again, I will care to reiterate over all of the threads that I have been posting in related to this...
This thread is a pretty good collection of all the places I have posted about this already (see first post)... http://forums.tweaktown.com/f31/diam...edition-27348/
I really want to hear some opinions...
No, even stock cooling has issues. That's the problem. You should try it on the stock cooled 4870. Unless your ambients are low and your fan speed is very high (like 90%+) running FurMark for a bit will most likely crash your card even at stock speeds. Even if you do not crash by some chance, running your VRMs to 125C+ is outside their operating range so you can expect failure any time. So far I have tested a bunch of 4870s trying to get to the bottom of this problem all with the same results. Changing VRM cooling though to something nuts like the mobo HR-09S Type 2 MOSFET sink stops the crashes and lowers the VRMs to a nice 80-ish celsius.