• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Help with analyzing video cards

Halecusis

n00b
Joined
Oct 7, 2012
Messages
6
Instead of starting a thread that only benefits me, (i.e. here are my rig's specs, what is a good card), I thought, why not try and ask a question that will benefit multiple people?

When I am doing research on my own, I can compare cards and see which ones have the highest numbers and come away with a rudimentary analysis of which is best. What I do not know, however, is what each of these respective "specs" affects or augments.

I realize that one aspect of a card may impact multiple areas, but I'm sure each aspect can be distilled into two or three primary functions, if not just one.

For example, what does having four gigs of memory get me that two won't? Does it raise your FPS? Is it for people that run multiple monitors? Should I only be interested in that if I am going to be rendering 3d models?

Doing my own research I run into two problems:
1. I haven't found a good comprehensive list of spec explanations
2. Most explanations are wordy and I have trouble understanding what having 1536 CUDA cores, or whatever, will do to my games and if I need that many.

Hopefully I have laid out what I am looking for in a way that one or multiple helpful people will slap a list of five or six important specs, and what the affect, together for us less than enlightened hobbyists.

Also, if this has already been addressed and I missed in my search a link would be appreciated.

Thanks
 
You want just enough memory so that your harddrive does not have to compensate for your lack of memory using virtual memory. That is what brings most old computers to a standstill, the harddrive light being on and windows being completely frozen until it can finish loading.

You can go with just 2GB now if you wish and still manage things, but for its lifespan I'd suggest at least 4GB.

CUDA cores; We used to increase the speed of CPUs until we hit a wall around 3Ghz-4Ghz where too much heat was produced at a higher speed. The only way to make CPUs faster after that was by having multiple cores on one CPU (2 or 4 cores running at 3-4 Ghz).

Graphics card have taken this concept and run with this-- and apparently it is much easier and more suited to graphics cards-- so you have 1536 CUDA cores on one graphics chip (!) all working in parallel to do computational tasks. I opted for the Nvidia 660 Ti over the normal 660 because of the big difference in number of those cores.
 
One problem with what you are asking is that the specs often don't necessarily compare well across generations or between AMD and Nvidia. So the number of stream processors matters when comparing the GTX 670 to the GTX 680 (where more is better), but it isn't such a great comparison between the GTX 680 and the HD 7970, or between those cards and the GTX 580 and HD 6970, because of architectural differences.

VRAM only matters if you run out - having more VRAM doesn't mean you get better frame rates unless you are in a situation where you are running settings that exceed the VRAM you have - in which case it makes a tremendous difference. However, for a single card, you will usually run out of GPU power before you run out of VRAM.
 
I would take anything written by THG with a grain of salt considering how sloppy their work are to put it nicely.

Anwyay, Forceman hit the nail on the head: You really can't compare based on specs alone. You can understand every single GPU related spec in the world and that still wouldn't help 50% of the time determine how fast a GPU really is due to reasons that Forceman mentioned. Two GPUs from the same manufacturers can have the same specs (VRAM amount, amount of shaders, clock speeds, bus interface, etc) but because one of those GPUs uses a different/newer architecture, that can and will throw off any specs based performance comparisons. Sometimes the specs are just fluff. Sometimes drivers aren't available that allows the game to take full advantage of those specs. So you really have to read and compare multiple GPU reviews, not specs pages, to really get a sense of how well a GPU will perform compared to another GPU when it comes to gaming. You have to look at the GPU as a whole, not just the individual specs.

So yeah, don't look at specs, look at reviews. That's a far more accurate way of determining how well a GPU really performs.
 
Last edited:
Thanks for the info guys.

I'm not surprised at all by the answers. I have always used reviews and benchmarks to make my decisions for GPUs. Although I have been putting together my own computers for years now, I am looking to become knowledgeable about what really makes things tick so I can make a computer sing.

I suppose it is comforting to know that I haven't been too far off on my methodology for determining the best GPUs
 
Back
Top