Reliable and longer lived video cards

carlmart

Gawd
Joined
Sep 17, 2006
Messages
691
After experiencing problems with two different brand GT640, one living less than a year and another just a little more than that, I am wondering if modern video boards are less reliable and short lived than old ones.

Is it being short-lived disposability a quality that might be now a part of computer stuff too?

The previous video board to them was a GT 9600, and it lasted 4 years, which seems like a better figure to me. In that time technology improves and you can get more by upgrading.

So I wonder what video cards you have been using in the past years, and which have lasted more. Preferably Nvidia types, as my video editing program does get along better with Nvidia.
 
The only NVIDIA cards I've owned that have physically died were 8800 GTX from BFG Tech, and it took more than 5 years for them to do so. I have owned a range of NVIDIA cards from entry-level bargain bin models to enthusiast models, and with the above exception I have only upgraded to keep up with technology.

I'm currently running a factory overclocked GTX 780 for going on 7 months, with a GTX 570 as a PhysX card that is on its fourth year in my system. Both cards are from EVGA.
 
I think you're just unlucky OP. I haven't had a GPU die in a very, very long time from either camp or any brand. GPUs tend to be more reliable these days, not less. I mean, certainly problems can happen, but in terms of RMA rates that i've seen - they're no different than they were in prior years. RMA rates have held steady at the 1-2% mark for many years now for all GPUs, so it's unfortunate that you fell into the 1% but I think things are better nowadays, or at least the same.

Now I will say that there are things you can do which can affect your GPU operation - over-voltage can cause electromigration, which can take a 10 year lifespan of a GPU down to 2-3. That can happen, but it takes time. But this isn't applicable in your case, sounds like you just had bad luck.
 
The last GPU I had that went bad was a GeForce 4 Ti4400... And that was well after I had upgraded it and gave it away to a friend.
 
It's all about heat. The cards that run hotter, for various reasons, will be the ones that usually die sooner. This is the main reason dual-GPU cards get a bad reputation for being unreliable, as they tend to run hotter than their single-GPU counterparts. Sometimes companies try to avoid a card getting the "dustbuster" label and set the default fan curve more toward keeping the card quiet than keeping it cool. Using something like MSI Afterburner to set a custom fan curve can help a lot in this regard.

It's not just about the heatsink & fan, some GPUs just run hotter than others.
 
It's all about heat. The cards that run hotter, for various reasons, will be the ones that usually die sooner. This is the main reason dual-GPU cards get a bad reputation for being unreliable, as they tend to run hotter than their single-GPU counterparts. Sometimes companies try to avoid a card getting the "dustbuster" label and set the default fan curve more toward keeping the card quiet than keeping it cool. Using something like MSI Afterburner to set a custom fan curve can help a lot in this regard.

It's not just about the heatsink & fan, some GPUs just run hotter than others.
What kind of temps are we talking about, and which components are we talking about?
I have a 4850 that idles at 85~90C since day one and it's still being used in an office rig today.

I load all my graphics cards at throttling point (usually around 90C) due to noise reasons and I haven't had any failures.
I can't see heat being an issue and believe me I've been trying.
 
Yes, I do think you're right. Even if I watched the card temp, I wished it did have some kind of alarm to tell me when it did went over a certain temp.

But now I'm thinking differently over this after something I did. As Zotac has taken over 40 days and not yet delivered the card they promised to, I found an all 7100 card I had with passive heatsinking. The problem is it couldn't stand any game that it got hot and turned the PC off.

So what I did was finding a fan that was not too small, which would be noisy, and used a rubber band to hold it to the heatsink. The result: even gaming won't increase the temp more than 40C. The only problem is it has a small memory and can't run more modern games.

So I'm thinking: why not get a modern passive heatsinked board, with more memory, and add a fan to it? Passive heatsinks are designed to properly dissipate heat, but they also get hot. What do you think?

Any board you would suggest?
 
EVGA has been rock solid for me for the last 8 years. Not one single issue with 4 different cards in that time.
 
It's all about heat. The cards that run hotter, for various reasons, will be the ones that usually die sooner. This is the main reason dual-GPU cards get a bad reputation for being unreliable, as they tend to run hotter than their single-GPU counterparts. Sometimes companies try to avoid a card getting the "dustbuster" label and set the default fan curve more toward keeping the card quiet than keeping it cool. Using something like MSI Afterburner to set a custom fan curve can help a lot in this regard.

It's not just about the heatsink & fan, some GPUs just run hotter than others.

Perhaps, but i'd be hard pressed to think this would ever apply to a GT640 though. The card isn't very powerful (in the context of 1080p gaming) and generally runs cool. Heck they put those chips passively cooled in ultrabooks...I mean the GT640 generally stays around 50C even at ridiculous workloads such as Unigine heaven, from what i've seen in reviews. So your statement may be true on a general basis, but it doesn't really apply to the GT640.

I could see that argument applying to some GPUs though. For sure TDP and power use has risen steadily over the past 8 years or so - it's pretty obvious that we're hitting a wall in terms of TDP, so there's a push towards efficiency now. Especially since mobile is such a thing - and GPU companies will use the same architectures for both desktop and mobile which will necessitate even further improvements to efficiency.
 
If you have this many video cards dying in such relatively short order, I would question the power supply and/or motherboard. I'm quite sure the load of a GT640 can be handled by almost any PSU, but that does not mean that it is getting clean power reliably from the PCIe slot.
 
The GT640 cost $100 at launch. The 9600 GT cost $190 at launch. The OP is comparing the longevity of the two and is surprised that the card that cost half as much didn't last as long.
 
I have owned:
Unknown Brand Riva TNT 2 (retired)
Dell OEM Geforce 2 GTS (retired)
Unknown Brand Geforce 4 MX400 (retired)
Abit Geforce 4 Ti4200 (retired)
Abit Geforce FX5600 (retired)
ATI OEM Radeon 9600 AIW (retired)
Saphhire Radeon 9600 Pro (retired)
ATI OEM Radeon x800xt AIW (retired)
ATI OEM Radeon X1900 AIW (retired)
EVGA Geforce 9600GT (in use in parent's computer)
Diamond Radeon HD4850 (backup for troubleshooting)
EVGA Geforce 560 (in use in HTPC)
EVGA Geforce 560ti (in use in GF's computer)
EVGA Geforce 760 (currently in use on main computer)

The only one that ever died was the Ti4200. It died due to my fault from a zap of static electricity on a cold winter day, low humidity, wearing socks on carpet, and my fingers were on one of the VRAM chips. I actually saw the tiny little lightning bolt come out of my finger. Hurt like hell for a second too and I dropped the card on the floor. :D

Every single one of those cards lasted 5+ years across multiple boxes. When they were retired from my main computers they were given away or put in back up computers until they were junked from simply being too slow. If the cards hadn't been trashed, I'm sure many of them would still be running. Right now, the longest currently used card of that group is the 9600GT, which is still running strong on my parent's computer. It is now a 6 year old card. Prior to that, the Geforce 4 MX400 was the longest running, as I installed it into a Pentium 3 backup box in 2008 which was stolen soon after.

-edit-
If you have this many video cards dying in such relatively short order, I would question the power supply and/or motherboard. I'm quite sure the load of a GT640 can be handled by almost any PSU, but that does not mean that it is getting clean power reliably from the PCIe slot.
My thoughts too, unless he is mishandling the cards like my incident with my Ti4200 listed above. However, I'm going to give him the benefit of the doubt.
 
Last edited:
Man, I have had 10 video cards in my history and none ever failed on me.

Maybe I'm just lucky?
 
The GT640 cost $100 at launch. The 9600 GT cost $190 at launch. The OP is comparing the longevity of the two and is surprised that the card that cost half as much didn't last as long.

Then I should look for something around $190?

The PS is a Corsair CX500, and the power line is dedicated to computer only. I have three separate power lines at home.

I don't know where else to look at.
 
After experiencing problems with two different brand GT640, one living less than a year and another just a little more than that, I am wondering if modern video boards are less reliable and short lived than old ones.

Is it being short-lived disposability a quality that might be now a part of computer stuff too?

The previous video board to them was a GT 9600, and it lasted 4 years, which seems like a better figure to me. In that time technology improves and you can get more by upgrading.

So I wonder what video cards you have been using in the past years, and which have lasted more. Preferably Nvidia types, as my video editing program does get along better with Nvidia.

I wouldn't say that modern video boards are less reliable. If you take an old video board and a new one, let them perform the same task, the newer one would be more efficient in doing it. Newer video boards are build to do more tasks and heavier tasks then older ones, which require more power and more heat is a biproduct of that.

Of the major, more widespread failures Nvidia have had with their cards the last decade when it comes to reference design, these 3 comes first to mind as examples:
1. The 'bumpgate" where high lead solder bumps was used on the GPU's which caused failure on many cards and also a lot of laptops. Lots of GTX 8XXX got recovered though by "baking" the cards.
2. Insufficient power phases on the first GTX 570 reference cards (if I remember correctly, there were only 4 power phases on those). I had fortunately bought a Gainward Phantom GTX 570 with sufficient power phases which lives well in a friends computer today.
3. GPU killing drivers (196.75 being the most notorius of them all):
https://www.google.no/#q=nvidia+driver+killing+gpu

Other then that, its been more the aftermarket cooling solutions that have been insufficient and not Nvidia's fault. Some scandals, (I think it was Gainward or Palit) which sent GTS 430's to reviews with heatsinks for the VRM's, while the retail version was without.

I believe its very rare that its the GPU itself that fails on video cards today. It seems more that its the memory and power distribution that fails (VRMs). Its not only about heat, but also about quality. The current and the heat goes up and down on modern graphic cards according to load. Many scuffs over the marketing speech of "military class" caps and chokes, but many knows that if you have quality caps and chokes, the chance of failure is much less. Some focus on VRM heat, but its not all equal when it comes to heat and the quality. Some can take much higher heat over longer period of time.

If you get a video card with quality power distribution and sufficient cooling, I belive that modern cards of today can have a long life. :)
 
OK, on the outside that board looks interesting, with a reasonable fan size and probably not too noisy.

Pity the EVGA is not available in my country yet, only PNY and Zotac. Quite more expensive than USA too.
 
That PSU should be able to drive a GTX 660 properly (well, mine does, and it's a 520W Compucase... thing). If you can get one of those instead of a 750 Ti for the same price, then I think it'd be worth the little bit of extra power. Mine is still holding its own at 1080p with a modest overclock, but I've only had it since last August, so I can't say if it's super-duper durable, though. However, it isn't a stretch to suppose that the low power draw and temperatures of the 750 Ti will lead to less stress and therefore higher lifespan, even if the 660 won't go warmer than 70 degrees in my case before instability begins (for some reason). Also maybe worth noting is the shortness of the 750 Ti's memory bus; 128-bit compared to the 660's 192-bit. Might have an effect on your program, but I really couldn't say.
 
Back
Top