Sign in with
Sign up | Sign in
Your question

XFX 4870 1GB acting up

Last response: in Graphics & Displays
Share
November 5, 2012 9:04:06 PM

Right so to start lets list my rig parts

OS - Windows 7 64-bit home premium (i'm poor)
Mobo - ASUS m4a78 pro
CPU - AMD phoenom II x4 950 3.0GHz
RAM - 8GB (4x2GB) OCZ gold 800mhz
PSU - Corsair 750W Professional series modular
GPU - XFX Radeon HD4870 1GB

Part 1:

My problem is that, as of last week, my card doesn't want to play at all, but only in my system it would seem. The computer can't display anything when using the card so i've no idea if it's loading the bios or not and i cant enter the bios to check anything because the card wont display anything.

On the few occasions the PC does display an image, it artifacts like a mother which is pretty crap and then the GPU fan goes 100% and displays a solid red LED on the D1601 which i am told is the critical temp fault. Now i know that's what it's saying but i know for a fact it's complete bull as in another system in the same environment (same ambient temps etc.) the card will idle at 45deg and stay bellow 70deg under full load.

My PSU is under a year old and i havn't had an issue with it at all however i'm still within european warranty of 2 years which is good.

Part 2:

Right so without the card inserted the computer runs perfect, without any hiccups at all. I can perform all tasks on the PC bar playing computer games or doing anything work related (3D architecture) and as such is non productive for me.

I must stress again that the system runs PERFECT without the card installed. I will be trying another PCIe card in my mobo to test the slot for a fault however the card is vastly inferior and as such cant put a stress on the PSU.

The good news is that i have purchased a 7950 GPU but cant get it till saturday so for now we can only speculate on the PSU.

Eveything has been removed and reseated appart from the CPU as i can assure you this is fine, im sure if it was a CPU fault i wouldnt be able to do anything at all within the windows 7 environment.

I have run memtest on my memory for a long period of time with an all clear result so apparently that rules the memory out. The only things i can see causing the issue are the PCIe slot, the PSU and possible the GPU (though it runs fine in another system).

I would also like to note that the XFX support team haven't helped me in any way other than tell me to spend more money on them, top class marketing.... i mean support team they have.

Please if anyone could chime in and provide me with some assistance into this matter as soon as possible i'd greatly appreciate it. I'm not behind the door with computers but this is driving me insane!

More about : xfx 4870 1gb acting

November 5, 2012 9:24:55 PM

Well it seems like you have tried everything but if you would just humor me for a bit. Have you tried cleaning out your registries and installing new drivers (Im guessing your using an AMD GPU so re install CCC)?. I usually do this when switching between Nvidia and AMD gpus but it might work here. If not hope the new card works out for you
November 5, 2012 9:57:52 PM

I've already used DriverSweeper to remove all traces of the cards drivers and CCleaner to clean the registry, the major problem is that i cannot boot with the card installed on my system in order to apply clean drivers.

I'm typing this on a laptop and my PC is on in front of me with no card in, just running the on board chip which is shocking but like i said previously, performance wise theres nothing wrong with it at all.

My system is around 3 years old now save for the PSU and my GPU wont run on it, however when it's installed in a 5 year old system (my dads) it runs without error. I'm pinning it on the mobo right now until i test it with an older gfx card but i cant really stress the PCU till i get the new card which is annoying.

It's obviously something conflicting but i cant see how the CPU or the RAM could effect it in way due to the RAM passing many tests and, well, if the CPU failed i wouldn't be staring a windows 7 desktop.

It could well be the psu, there are a couple of things i need to try with that yet, i have some molex adapters for the PCIe 6 pin power inputs that need to be tried, if that works then the PSU +12v rails are faulty, if that doesnt work then it's still either the mobo or the PSU failing.

Today i took apart the GPU and thoroughly cleaned it, applied new thermal paste in the form of Arctic Silver 5 and reseated it to be greeted with the windows logon and desktop for roughly 10 minutes before i had a 'critical temp failure' at 45deg C, which seems like a false failure.

Some more info here, when i press the power button to boot, the card tries to post twice instead of once like it normally would, so you've got the fan running at 100% on 2 short intervals to then display nothing however all the components sound like theyre working fine (the card even sounds normal) but with no visual feedback it's hard to say.

For now i'm attempting nothing till tomorrow evening (work you see) and its nearly midnight but please fill this thread with additional info and help, i'll try all suggestions if not already performed.
Related resources
November 5, 2012 11:10:47 PM

In order of RISK..

1. Try the catalyst 11.12 and app sdk 2.4 for open cl support. The newer drivers drop support for open cl. I found this older driver to work better with the hd-4870.

2. Try unplugging and replugging in the card into the pcie slot about 20 times to clean the contacts, it could just be a dirty contact. Also do this with the power plugs about 10 times.

3. Tie the back end of the card up so it lays flat, sometimes they sag in a tower case causing cooling issues.

4. Take the square metal spring that is on the back side of the card off, the one with 4 legs on it.

http://www.ixbt.com/video3/images/rv770-2/his-4870-scan...

Cut a plastic shim from something like a coffee container lid, margarine container lid,
so that it will be under the metal square but clear the 4 mounting screws with the tiny springs under the screws.

What happens is the metal spring looses some of it's tension. Also the dabs of silicone or glue on the bottom of it gets compacted allowing the metal square to contact the card in places.

Start all 4 screws with a few turns first. Then tighten them a few turns each at a time in a cross pattern, so that they get tightened evenly and not all at once.
If you tighten one screw without the others you could crush the gpu core.

I fixed a couple of cards this way. One that was totally dead and now works fine.

PLEASE NOTE: THERE IS SOME RISK WITH THIS REPAIR ATTEMPT. YOU COULD TOTALLY CRUSH YOUR GPU CORE AND KILL THE CARD FOR GOOD.

DO THIS AT YOUR OWN RISK!!!

5. BIOS MODDING AND FLASHING CAN KILL YOUR CARD, AGAIN THIS IS AT YOUR OWN RISK!!!
You will need the card to be at least working in 2d mode to be able to try this..
This might fix it if the card only crashes when you run games or 3d tests.
http://www.techpowerup.com/downloads/Utilities/RBE
bios modding tool..
www.techpowerup.com/gpuz
copy your current bios for modding..
http://www.techpowerup.com/downloads/2107/ATI_Winflash_...
flashes bios..

Things to try with the RBE editor.
You could try changing the fan temp profile so the fan keeps the gpu a bit cooler when idle and under load. You will have to find a compromise between fan noise and temps.
Dropping the default startup gpu speed to 450mhz 2d and 700mhz 3d mode.
Dropping the memory speed to 800mhz from it's stock 900mhz.
Increasing your card's gpu and memory voltage if the card allows it.

November 6, 2012 3:03:19 PM

Posting from my phone, more of a quick update.

So I'm leaning toward a PCIe slot issue at the moment, I've just got home and plugged my dads HD 3600 card in, it booted fine however when I tried to run FIFA 13, chosen as my dad runs the game on his PC with this card, but the PC shutdown instantly. This problem hasn't happened without a card installed, fifa runs quite nice with onboards and I suppose would look ok on a 5 inch screen.

Again I can't test the PSU yet. I did read that CPU heat fault can cause a total shutdown but like my GPU the temps are incredibly low, the bios shows CPU idle at around 30deg C which I would say is ideal.

Ill do some more playing around with this card and see what happens but I can sense some common ground between the problems, seems strange that a system would crash only when there's a PCIe card in.

By the way, need4speeds, thanks for that massive list! It should keep me busy all week! Ill keep you posted on the progress of those.

I don't know how much of a spanner my new finding throws into the works but anything is worth a go at the moment. I would like it to be the mobo, mainly because on reflection it was a poor choice.

Anyway ill continue and get back with any new info.

Again if anyone thinks of anything extra just post and ill try it out.

Cheers.
November 6, 2012 3:10:26 PM

Another quick one, tried fifa again immediately after posting but on the lowest graphic settings, as soon as the game required 3D graphics (the menu where you would select game mode etc.) the system powered down again.

With this knowledge is be willing to bet my 4870 is fine, which is good as I can pass that on to somebody else.

Going to try the onboard graphics again and see what happens this time.
November 6, 2012 5:55:23 PM

I understand putting links into a thread from another website is against some forums rules but I just came across this page whilst looking for others with similar problems

http://superuser.com/questions/440293/pc-powers-off-at-...

Now I don't know how similar this is to what in experiencing but, just like this guy, my PC will power off at any time, no matter if I'm in windows or in the bios. Our biggest difference here is that I'm pretty sure mine is down to the PCIe slot.

Played with 2 card a now that I know work in another machine and both have resulted in a system power off. seating nothing in the PCIe slot works fine. But we can rule out the PSU now.

The guy in the link I just posted resolved his issue with a new motherboard, not a cheap route for me sadly, means a new board and new DDR.

I do have a theory though, it may sound stupid but hear me out. Is it possible that connecting a PCIe card is causing some sort of short up to the CPU? I might discover a short when I BreadBoard the components, this will be the last test before any risky tinkering! (Although I doubt my findings will differ)

Sorry for the triple post by the way but for some reason I can't edit from my dog and bone (phone), unless that's just this forums rules? Correct me if I'm wrong.
November 6, 2012 7:52:20 PM

It's not the card then since the trouble doesn't follow the card. Do try another power supply if you can.

1. It could be the way the board is installed in the case. Take the board out.
Look under where the pcie slot is on the case, look for any mounting lugs that might have been missed that might be in the wrong spot and shorting out the board.
If there is nothing, maybe the board sags and touches the metal case under the pcie slot when a card is put in that slot.
Take a piece of cardboard or paper, fold it so it's about the same height as one of the board mounting lugs. Place it under where the pcie slot is on the metal case, reinstall the board. You will have to try the paper/cardboard a few times to get the right "spring" to it. Foam earplugs also work well or just a piece of foam.

I have seen this happen before.

2. It could be a dirty atx power connector or the 4 pin power connector on the board.

Try plugging and unplugging them a few times and retry it.

Look at the atx and 4 pin connector for any tarnished or burnt connector pins, normally the plastic beside the pin will be melted a bit.
If you do find a few bad pins, carefully squeeze the pins a bit so they plug on a bit tighter. (the female pins on the connector, not the male ones on the board)
There is a fine line between crimping them not enough and too much and they close and wont plug back onto the board. It's pretty hard to fix this.
If the power supply is under warranty, do nothing and send it in for rma as this repair voids the warranty.

November 7, 2012 7:13:29 PM

Just doing a bench test currently and it's all working fine, graphics card is playing games under full load without hiccups but its still trying to post twice which is strange. As good as this sounds I'm very unhappy as it means yet more testing, joy.

The case is now empty save for additional HDDs and optical drives and I've taken your advice and had a good look around, no visible signs of the motherboard pressing to the case (scratches etc.) but I have found some longer stand offs for the mother board that ill be using when I rebuild.

I had the bare minimum in when bench tested and I plan to rebuild again with the bare minimum and add things (I've done this already but I'm not sure anymore).

Started with 1x2gb mem stick which threw me at first, so I switched off and inserted all 4 sticks and it still worked which confused me again.

Now what I'm thinking is although I feel the PSU is fine, maybe it's hidden it's fault well from me, so ill be re-adding to the computer every boot to see how she goes, this will take all night so unless I'm still awake don't expect an update about it.

My new guess is that for some reason, although the PSU is more than capable on paper, it suddenly can't handle the load and as such is powering down.
Ill just add to this that I'm using an NZXT case where the PSU sits at the bottom of the case. It's getting good airflow all round the outside of the case from what I can tell.

I shall return, thanks again for the further advice
November 7, 2012 8:31:02 PM

You cant change out for longer standoffs. The part where the card screws down will be also be higher. Also the board will sit higher in the case and now the top of the board might hit something like the hdd or cdrw racks. It could create a whole new problem.

I suppose you could bend the card's metal tab, but that is not the way it's supposed to work.

What i meant is some boards will flex down and touch the case. This is fairly common.

It does sound like this fixed something because it was almost dead before.

I would try and find out why it is now working.

If it only shows up for a while and then returns to a single catalyst icon, the card should still work ok like that. I have seen this before, it's some glitch with catalyst. It never seemed to cause any troubles with my cards.
November 8, 2012 2:33:16 PM

Yeah I found out about the stand offs while I was fitting it, sempt like a good idea at the time! I did feel silly but it's starting to exhaust me a bit.

I've got it in the case, have only 5 stand offs now, one in each of the corners and one in the middle right position to help with the DDR, the slots seem a bit stiff so the board tends to flex but its rare they ever come out, only to clean or troubleshoot. The GPU actually slots in fairly easily in this mother board, but I do understand what you mean with the flexing, it can be a pain and its slightly bum twitchy when it starts going too far.

Got a few more things to add to the PSU today, got up to having all HDDs connected, GPU, CPU and of course the motherboard. Ill be adding the disk drives first and then the fans. I don't know the possibilities but I had a thought that one of the fans could be causing a short in the case. So hopefully we might have an answer by tonight, if I connect it all and it still works I think I may have to put it down to magic. Yes, it's sending me slightly mad.

I'm going to power up first and let it run a while and see if the 'fix' is carrying across today, I did 7 boots last nights and didn't have a single hiccup save for the card doing its post test twice on some occasions. No graphics errors though and no sudden power downs so its back to normal for the moment.

I'm trying to stress the power as much as I can right now because if the PSU has gone faulty on me ill be able to return it on Saturday when I collect my new GPU; although the 7950 does sort of look like it might eat the rest of the PC....

Again ill update with any more info I get from this marathon of a troubleshoot!
!