• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Help diagnosing IceQ HD 7950 issue

vanp

vanp

vanp

Associate
Joined
8 Oct 2011
Posts
20
Hi Chaps,

In light of the GPU price madness at the moment, I'd like some help in diagnosing an issue with my IceQ HD 7950 as it's still a really decent card (or should be).

The problem I'm having is that it refuses to operate at anything more than PCI-E 2.0 x2.

The card was originally my brothers and he had a similar issue with it.

It will play a lot of A - rated games and slightly older A-rated games, perfectly fine at 2560x1080 resolution - I have an ultrawide monitor (at the aforementioned speeds) for example, GTAV runs perfectly well, as does Dota2, Civilization 6 on full graphics - but it's starting to come unstuck on bigger games, in particular, I just installed Battlefield 1 and on the small maps it is fine on medium graphics, but the larger maps are almost unplayable.

As I understand it, from various internet research, the 7950 is still a decent card and should handle BF1 on high settings perfectly fine @1080p, without lag or huge fps drops. So I think it is getting massively bottlenecked by the x2 bandwidth.

Any ideas? My old HD6870 runs as it should still, and the 7950 did the same thing in my brothers PC, so I know it's not a motherboard issue..

Just in case you need my spec:

AMD Phenom II 1100T X6 Black edition
16gb Ram
840pro SSD
 
I've heard of the following case before, unsure if it applies to your case, but it's fairly easy to check :) :

Do you have another PCIe card in another joined GPU PCIe slot? eg. You have the 7950 in your top PCIe x16 slot, but you another another card that's been plugged into another x16 slot and this other card only goes at 2.0 x2, like a TV Card or Raid Card or something. This could cause your described issues as the linked PCIe slots goes at the slowest rate of the cards connected.
 
How do you know it's definitely running at Pci-e 2.0 x2?

Worth checking with a couple of different programs and/or changing bios settings e.g. force slot to pci-e 1.0 (if you have the option) to make sure that what is being displayed actually changes, and isn't just a software issue.
 
Meddling-Monk said:
I've heard of the following case before, unsure if it applies to your case, but it's fairly easy to check :) :

Do you have another PCIe card in another joined GPU PCIe slot? eg. You have the 7950 in your top PCIe x16 slot, but you another another card that's been plugged into another x16 slot and this other card only goes at 2.0 x2, like a TV Card or Raid Card or something. This could cause your described issues as the linked PCIe slots goes at the slowest rate of the cards connected.
Click to expand...

Thanks, I'll check that tonight as I do use an Asus Xonar DGX PCI-E card, although I'm fairly certain it isn't in 1 of the main PCI-E x16 lanes.

Armageus said:
How do you know it's definitely running at Pci-e 2.0 x2?

Worth checking with a couple of different programs and/or changing bios settings e.g. force slot to pci-e 1.0 (if you have the option) to make sure that what is being displayed actually changes, and isn't just a software issue.
Click to expand...

When troubleshooting with my brother awhile back we tried GPU-Z and it confirmed the same. It's the AMD Radeon Crimson driver suite that is reporting it currently at that speed - it doesn't move during gaming from what is written. Also as mentioned above, the performance is so woeful at times, it must be running at the speed it says it is.

Any idea what to look at in the bios that could affect this?

I have read something about a Turbo_USB setting, which I will check tonight.
 
*Update*

Checked bios settings, couldn't find anything that would hinder the card

Checked that no other PCIE card could be stealing bandwidth = same result, max PCIE 2.0 x2 speed

Used GPU-Z and the render test as well as the Crimson Suite, both report that it runs at only x2 under load.

Realised that the card was actually in the 2nd PCIE slot and tried swapping it into the main slot = card won't display at all (then remembered this is why I put it into the 2nd slot)

Then tried my older HD 6870 in the main slot to double check = card runs as it should in the main slot @ PCIE 2.0 x16 speed

Had a look at the bios repository and it has the latest bios for that particular card

Anyone else have any bright ideas? The HD 6870 is noticeably slower at max settings for games such as Dota 2, even on x16 mode as opposed to the 7950 only running at x2 (I would have thought it would struggle a lot more at only x2) - so for now I guess I shall just put the 7950 back in and bend over and take it until Graphics cards prices come down.. :rolleyes:
 
so the other slot on the board is dead? could be a potential issue with the board, you'll only figure that out if you can see damage on the board by the slots or by trying the 7950 in another pc. It could still be the card though, look for damage across or near the edge connector on the bottom of the card. Make sure no tracks are damaged ect. Also, take the cooler off and look at the paste on the gpu - i had a 7950 that started crashing to desktop and grey screening constantly and after a lot of head scratching it turned out the paste was slightly conductive, or if not that then certainly capacitive, and it was all over the SMT caps next to the core (not my fault, i bought he card 2nd hand...). One clean up job later and it was solid as a rock again. You never know what might be affected when you get thermal paste on components lol.
 
No I think you misread - the card was like this in my brothers PC and it works and displays an image and runs at x2 in the secondary slot but not the other. However, I tried my 6870 in the primary slot and it works perfectly, so there is nothing wrong with the motherboard - the card must be at fault. Will consider taking it apart and checking over the paste etc.. but the card runs absolutely rock solid, no crashing - just at x2 speed
 
That sounds rather suspicous. What motherboard do you happen to be using?

:: edit ::

Just want to look at lane palcement is all, not directly related to the issue at hand or blaming motherboard.

But I would surmise that there's an issue on the PCIe connector then on the card wiht what you've said then, but this will be a physical fault issue. And I'm sure it's out of warranty at this stage.

Incidentally, do you have another x16 lane slot open on that motherboard?
 
Meddling-Monk said:
That sounds rather suspicous. What motherboard do you happen to be using?

:: edit ::

Just want to look at lane palcement is all, not directly related to the issue at hand or blaming motherboard.

But I would surmise that there's an issue on the PCIe connector then on the card wiht what you've said then, but this will be a physical fault issue. And I'm sure it's out of warranty at this stage.

Incidentally, do you have another x16 lane slot open on that motherboard?
Click to expand...

I'm using an Asus M5a99X EVO R2.0 - with the latest bios version

It has: 2 x PCIe 2.0 x16 (x16 single or dual x8)

Unfortunately it is out of warranty. As mentioned before, my brother had exactly the same issue. I was speaking to him about it last night and he said he bought it from Overclockers, he raised the issue and it went back for testing and then eventually went to HIS (the manufacturer) - eventually it got returned to him with 'no fault found' Now I don't have a clue how they came to this verdict, seeing as it is easily replicable. But I think he basically got annoyed with the months of waiting and as he wasn't playing demanding games, he just lived with it. But he said he has tried it on 2 different motherboards before it came to me and both had the same issue.

Any idea what exactly could be done if it is an issue with the PCIE connector.. is there somewhere these can be repaired? I'd attempt it myself if I had a guide or something to follow..
 
Well, if it's gone back to the manufacturer before, most likely it won't be it. But give what james.miller suggested a go, as it seems to cover the same ground I'd be suggesting to do at this point just to eliminate any remaining possible issues or defects that an end user can try and do for a recovery.

So thats:
Visual inspection of the gold connectors on the PCIe connector, make sure all are there (one will be shorter than the rest, this is normal, but again, most likely should be OK as it was sent back, but do it anyway just in case someone wasn't doing their job back then)
Check for scratches on the etched tracks/circuitry.
Remove old paste, add new paste.
Check for dust, clean dust off.
Check nothing wrong with PCIe Power Molex connector.
Check windows is not doing some power limitation to the GPU by reducing it's speed. (Last one I can think of that's not listed so far)

If nothing helps, I'm afraid it's just a matter of using as is until you get a new card.

But make sure whatever you do, is to always do thorough checks with the new card and if anything wrong with it, don't let up until it's fixed.
 
*Update 2*

Just got in from work and took a much closer look at the circuit board for the GPU this time.. I think I've found the culprit (please see exhibit 1, where I've ringed in red)

2872omx.jpg


I guess that must be the culprit then? A little bit of circuitry is missing, is it a bridge for the pins? So frustrating that something so small could have caused this and very annoying that the HIS 'technical' (ha.) team didn't spot it.

Is there anyway to fix this? Could I bridge it by using the tiniest amount of solder between the 2 points, or does it need whatever that tiny square block is? :rolleyes:
 
I would agree that it would be the likely culprit.

As for repair, that's a bit beyond my current experience. Maybe one of the more experienced forum users here may be able to direct you better, otherwise you'll only be getting my amateur info. But from what I've been reading about, that should be something that a user can repair.

:: edit ::

Oh, I think solder won't be enough on its own. I believe the missing piece may require replacing somehow, but that's as far as I'll go on for now. Again, hopefully someone more knowledgable can direct you further.
 
vanp said:
*Update 2*

Just got in from work and took a much closer look at the circuit board for the GPU this time.. I think I've found the culprit (please see exhibit 1, where I've ringed in red)

2872omx.jpg


I guess that must be the culprit then? A little bit of circuitry is missing, is it a bridge for the pins? So frustrating that something so small could have caused this and very annoying that the HIS 'technical' (ha.) team didn't spot it.

Is there anyway to fix this? Could I bridge it by using the tiniest amount of solder between the 2 points, or does it need whatever that tiny square block is? :rolleyes:
Click to expand...

From my limited understanding that is going to be tricky to repair, depending on what equipment you have at hand and how brave you wish to be. The good thing is that you have one right next to it most likely will be of similar value.

First you need to find out if its a resistor or a capacitor and also working with surface mount is going to be tricky unless you have experience. if its a resistor you will need to find out what resistance and package type. If you have a multimeter you will have to remove the other resistor next to it first and then measure its resistance and then you will need a set of calipers to measure the length of it to get an idea of package type.Then order 4-6 of these thing if you don't have experience with surface mount because you will likely botch it a few times.

http://www.interfacebus.com/resistor_table.html

if its a capacitor then you will have to measure its capacitance and this will be made easier with a oscilloscope which has function generating capacity along with a resistance of known value and there are other ways of measuring its capacitance or you can buy a capacitance meter just for this which would be easier if you don't know much about electronics.

The easier thing would be if there are any marking on there, like numbers or letter that you can read, please type there here and it should be able to identify it like that.

the easier way to mount the components would be to use solder paste and a hot air gun or you can also use a soldering iron if you have done this stuff before.
 
It's going to be a cap, yep. Id look at buying another card off the bay, dead preferably, and taking a similar cap from that. All those caps by the edge connector are likely to be of the same value or something very close and any cap you put on is going to be safer than bridging it (although I'd be tempted to do that ....)
 
Last edited:
Value of that coupling cap won't be vital, probably anything you can get in that package size ("0603" - 1.6 mm × 0.8 mm or "0805" - 2.0 mm × 1.25 mm, get your calipers out if you have any ;)) will work to make electrical connection. If replacing I'd look for 100nf 0603, cheap as chips on the bay etc.
 
Thanks for the replies guys - I have a soldering iron and have a little experience with it - unfortunately it's massive and I reckon anything I attempt with it would just get a massive splodge of solder on it, probably crossing multiple components :p

This cap is really tiny, I think smaller than the sizes mentioned above, maybe 1mm max in length, maybe 0.5mm across, no more than that.

If I were to attempt to solder between the 2 points, what is the worst that could happen? I understand a capacitors job is to keep the +/- charges separate, which a blob of solder obviously isn't going to do..
 
Even smaller would be 0402... Reach for the magnifying glass :D

It's there to smooth the voltage between the card and the slot. I'm a bit rusty but I believe it's part of the PCIe spec to ensure noise/voltage spikes is reduced between the board and add in card.

Like i said personally I'd bridge it and see if it boots x16, then worry about replacing it... Maybe.

I use a fairly large iron for everything, it's better to get in and out quickly (transfer the heat) on PCB SMD stuff than burn components/traces trying to heat up the pads and flow the solder.

To replace it I'd tin both pads with a TINY bit of good lead-tin solder, place the component over the pads and hold in place with a pencil, apply the iron tip to the whole lot in one go and pull it away as soon as the solder flows, holding the cap in place for a couple seconds until set
 
You must log in or register to reply here.
Back
Top Bottom