Upgrading graphics card on old CPU - makes sense or not?

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

How do you mean it "burned" like heat damage? Because the driver has no barings on the cooling of your system? But if you just mean the performance, then its more than likely your manufacture used its own "custom" BIOS that was incompatible.
 
Actually, it was the fault of the drivers. What they did was accidentally disable the GPU fan after installed. It led to an epidemic of GPU failures due to GPUs overheating when their fans were disabled. Look it up if interested.
 
Is that the same driver release that started to allow you to OC your video card in the control panel? If it is, then it was not the driver, it was the BIOS of the video card, look it up 😀. The driver released a feature that the BIOS was not compatible with. What happens is PNY or Gigabyte or whoever, thinks they can do better than NVidia and tweaks the BIOS to meet their "settings" when in fact, they change something that NVidia has plans for down the road. THUS the reason, they tell you to go to the manufactures website to obtain a driver, not the "generic" (even tho it is a much better performing driver release) so it is not "incompatible"

This is another big reason to go with big name brands.
 
Drivers are simply software that control how the device operates. Up until just recently (2 years ago) the fans were controlled NOT by the driver, but by the BIOS itself, otherwise when you do a clean install of Windows your fan would not work at all until you install the driver, which for some people is never.

Now the BIOS of the video card has its own "onboard" (don't know what else to call it) way of managing the fan speed. But when NVidia released its control panel where you could OC your video card using the sliders, it gave you the option to change that for the fan as well. Well some of your manufactures (I believe EVGA was one of the major ones to be hit) added their own code to the BIOS to manage the cooling of the video card. So when NVidia released their Driver to control the fans, their code was not there, so it didn't have anything to control.

PS refrence cards are not released to the public, they are only used for testing inside the company. No one would have any info on if they were affected.
 


Perhaps we're having a misunderstanding of what reference means. From what I've been told for many years, a reference card is a card that uses a stock PCB and cooler as was designed by Nvidia with clock frequencies and such that are also unmodified from the Nvidia design, IE a generic card. We call those reference cards because they abide by Nvidia's specifications and all of the major sites and more use this terminology. We call the cards that are not released to the public engineering samples.

Also, I really don't care exactly how the problem happened at this point. Nvidia said that it was their fault back then and never changed that statement. Perhaps it truly wasn't their fault, but that doesn't change the fact that they were related to the problem and their driver update burned many cards' GPU.
 
reference cards do abide by NVidia's specs, and are unmodified. Lets take a look. We will look at first, the reference design for the 650 video card (just the first one that popped into mind)
This is the reference design: http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-650/specifications
Notice it has a clock speed of 1058Mhz and 1GB of VRAM.

This is EVGAs card: http://www.evga.com/Products/Product.aspx?pn=02G-P4-2651-KR
Notice the same clock speed, but different amount of VRAM. This is not, obviously, a reference design.

This is Zotac's model: http://www.zotacusa.com/geforce-gtx-650-zt-61001-10m.html
Notice the clock speed, it is different than the reference design by NVidia.

And EVERY manufacture is different. The "reference" designs are simply that, for reference. The manufacture will make them vary to their will. Now I have only showed you the actual specs of it, I didn't show you the code for the BIOS. I had a PNY 9800GT XLR8 edition, and I COULD NOT flash my BIOS with a EVGA brand or anything like that, it had to be the XLR8 edition, or it wouldn't work. (it was also about twice the size of the EVGA BIOS) Obviously a more complex code. Also, you can click on the product images, and see a picture of the reference card that NVidia sends their manufactures. And as you can see, no card on the market looks like that.

Also, if you pay attention, when you FIRST start your machine your video card BIOS pops up first, it will tell you the brand, and model, and the BIOS version. I almost guarantee that no other card you will see is like that. I got 2 9800GTs at the same time, and they had different BIOS revision (thus the reason I know so much about it, via extreme troubleshooting). They would not run in SLI right because the BIOS revision was WAY off...
 
Each video card company may have some non-reference models, but just about every company sells a reference model for each card that they utilize in non-reference models too. For example, every major company such as EVGA, Sapphire, Gigabyte, MSI, and more all sell some reference cards.

Regardless, IDK why you're going off on tangents. The exact causes of the issue don't change the fact that updating to that driver version burned many GPUs and that was a problem that shouldn't have happened nor do the exact causes of it change the fact that Nvidia went out and said that it was their fault.

Furthermore, even if the issue was caused by straying from Nvidia's design (which it seems to have not been since the issue affected reference cards as well as non-reference cards), Nvidia should have made sure that the issue wouldn't happen. Updating to current drivers, even if they're not from the card manufacturer's site, doesn't cause the GPU to burn. Sure, they might not work, but they don't cause the GPU to burn by disabling the fan or any other way.
 

It happened because you got the driver from Nvidia, not from the manufacture. The manufacture wouldn't have givin a driver with that issue (cost them money in RMAs)
 


It happened to reference cards too. If Nvidia's driver doesn't work on reference cards, then what's the point of Nvidia posting their driver anyway? Furthermore, why does this issue not happen with subsequent Nvidia driver releases?
 

We just went through this. There would be NO public data on a reference card. NVidia is the only company that makes a reference card, NVidia DOES NOT sell cards to end users. They GIVE the cards to manufactures. I would be surprised if ANY video card is 100% like NVidia's reference card, hell Ill even settle with 90% like a reference. With that said, NO ONE should be downloading the drivers from NVidias website. You should do so from EVGA's website, or PNY's website, or someone else...
 


Actually, Nvidia does in fact sell their own cards, granted they're usually only sold to OEMs. There is plenty of public data on reference cards too because we can all look at the exact specifications. We can even look up the board layout of the reference PCBs and the specifications of the reference coolers.

Furthermore, just about every video card company has purely reference cards. All cards (at least in the last few years) have had many reference Nvidia and AMD/Ati units sold right on Newegg and other such sites.

I asked why Nvidia posts their drivers if we're not supposed to use them, not should we use them or not. I also asked why is it that their releases after 196.75 did not have the issue that 196.75 had if that issue was caused by the driver looking for code that wasn't there since that data shouldn't be in the video card's BIOS from a driver update after 296.75 either, should it?
 
I've got two Nvidia-branded GTX 295s as an example. Like I said, they're usually not sold except in some OEM machines, but they do exist. Reference cards with other brands are all over. For example, Newegg has dozens of them.
 

I own my own computer repair shop, and been doing this for years. And I have only seen 1 reference card, ever, and it was at a conference.
 
They're everywhere. Many of the cards that I've used were reference cards, although I usually use above-average non-reference cards because reference cards tend to have poor cooling. I build and sell computers for clients and I've probably built hundreds of systems.
 

well I guess I just don't see them, and I guess they are never benchmarked, or anything
 
Tom's benchmarks some of them and TPU benchmarks tons of reference cards. For example, I think it was Tom's first benchmarks of the GTX 690 that used a reference model. Tom's also benchmarks a few engineering samples (for example, Tom's benchmarked a Radeon *7850* single slot card with a 768 core GPU), so maybe they're not the best example, but most review sites benchmark some reference cards. Anand is another good example of a site that benchmarks many reference cards. Guru3D does too, but they're AMD benchmarks are often useless because they don't like to use up to date AMD drivers for their tests.
 
I was referring to a actual benchmark... like if you went to 3dmarks website and looked at top results. Im trying to say, no normal person has one. So NO normal person should be downloading drivers from the Nvidia website, they should get them from their manufacture, which is why this whole thing started.