Graphics cards: when is an upgrade not an upgrade?

[OldProf]

I don't even want to think about how many graphics cards I have "upgraded" to in attempts to run various incarnations of TS better.

For several years, I've been using an EVGA GeForce 2GB GTX 550Ti with reasonably good results, but there's always the temptation to move on up, isn't there? So when I saw a sale at Amazon for a GeForce GT 740 4 GB (single slot model), I figured that the higher number plus an additional 2 GB memory had to be better. The card arrived, I installed it, and I really cannot see an improvement in terms of fps or graphics in general. The old GTX 550 Ti required a power connection, but the new GT 740 does not. Since I can return the GT 740 easily (blessings on Amazon), I just ordered a double-slot version of the same card that does require a power connector.

Thoughts? Am I going around in circles? perhaps even backward? Will Santa bring me better knowledge about any of this for Christmas?

[GSkid]

I've heard people have this issue when they upgraded a CPU, but I can't remember hearing this issue with a GPU upgrade. I imagine the 2 slot version is for a bigger memory channel and bandwidth.... not to mention the bigger heat sink and power usage..

I think I may know your problem....your old card and new card are virtually the same in performance according to GPUBoss.com ......

Here's your match up.... just click to see the stats.... REMEMBER TO SCROLL DOWN THE PAGE TO SEE ALL THE INFO....

http://gpuboss.com/gpus/GeForce-GTX-550 ... rce-GT-740

Actually..... your old card has better Passmark scores.

That's actually one of the least detailed comparisons they offer. Other card comparisons (both desktop and laptop GPUs) usually include more tests and info.... for an example... I'm returning my new quad core i7-4710HQ w/ Intel HD 4600 GPU laptop after all and getting a dual-core i7-4510U (4 threads - 2GHz base/3.1GHz turbo) and the 4GB Nvidia GTX 850M. Here is a head to head of my current card VS. the new one I'm getting....

http://gpuboss.com/gpus/Radeon-HD-7670M ... e-GTX-850M

My match up has more info. Most importantly in this case is the 3DMark06 score. That measures DX9 performance....Railworks uses DX9. According to that score, my GTX 850M should have roughly 4x the performance (1,061.5 Vs. 4,452) than I do now. Could be more or less depending on other factors. Passmark is 829 Vs. 1,445 which implies almost 2x the performance. We'll see when I get it.

Also with Nvidia cards, the "GT" prefix is their budget line..."GTX" is their mid to high performance line. Your new card has a higher model number, but cuz it's a budget GT version.... in this case.... it's performance is identical to the GTX version of an older, lower numbered card.

Just a tip when using GPUBoss. When comparing GPUs, just enter the card's NUMBER and the options will pop up for you to choose. Typing in Nvidia or GT won't work as well or as quickly. Once you compare, you can also click on the card's name to get the info page dedicated to that GPU.

If no 3DMark06 score is available, Benchmarks, Compute Performance and Passmark in your current match up is all you have to work with.

[DrewG]

If your looking for an upgrade look at GTX 970's and 980's. Lots of power and on new egg some 970's are listed as low as 400, which would probably go for less during the Christmas sale.

[Ericmopar]

Tom I did a little research for you.
You'd be looking at going to a GTX 660 or 760 for any significant performance increase. The GTX 660s are still available for about $150.
The GTX 760s are available sometimes in that pricpoint, but in less desirable makes, like Zotac.

Generally the best card makers are MSI, Asus, Gigabyte and EVGA.
I could help you more, if I know your power supply wattage and what your spending limit is.
Also, what is your CPU?
Eric.

[buzz456]

I have a GTX650ti running on a 475 watt and the performance is very good for TS2015.

[OldProf]

I've heard people have this issue when they upgraded a CPU, but I can't remember hearing this issue with a GPU upgrade. I imagine the 2 slot version is for a bigger memory channel and bandwidth.

I think I may know your problem....your old card and new card are virtually the same in performance according to GPUBoss.com ......

Here's your match up.... just click to see the stats.... REMEMBER TO SCROLL DOWN THE PAGE TO SEE ALL THE INFO....

http://gpuboss.com/gpus/GeForce-GTX-550 ... rce-GT-740 (... etc.)

Thanks for the site reference, but the cards they compared each had only 1GB of memory, whereas my GTX-550-Ti had 2 GB and the new GT-740 has 4GB. Doesn't the amount of memory have an effect on a graphics card's overall performance?

[OldProf]

If your looking for an upgrade look at GTX 970's and 980's. Lots of power and on new egg some 970's are listed as low as 400, which would probably go for less during the Christmas sale.

Believe me, there's no way I'm going to spend 400 dollars on a graphics card, much less over $1,000 on that card someone else mentioned!

[GSkid]

Thanks for the site reference, but the cards they compared each had only 1GB of memory, whereas my GTX-550-Ti had 2 GB and the new GT-740 has 4GB. Doesn't the amount of memory have an effect on a graphics card's overall performance?

The quick and easy answer is NO ... or very little in most cases depending on the resolution and what filtering effects you use at those resolutions. So the performance shown for your card on GPUBoss is basically the performance either memory config should perform. They usually choose a standard resolution, graphics quality and filtering option that all graphics cards can easily handle so that it's a fair and universal benchmark to measure with. Settings that even a 1GB card can handle with little problem. I think that's why some game specific benchmarks are done at a "very high" setting instead of the "ultra" setting because the 1GB cards can't handle that particular game at that setting. Most likely from the very high res textures and the large amount memory they take up at a particular resolution (1920x1080 the standard bench resolution these days).

Some tests aren't affected by memory size at all. They are of the actual GPU's horsepower so to speak.

Here is an excellent Feb 2014 Tom's Hardware article basically answering your questions on video memory and other factors called "The Myths Of Graphics Card Performance: Debunked, Part 1"..........

http://www.tomshardware.com/reviews/gra ... 694-5.html

It's a good read and sorts out the facts from the myths on this video card stuff.

[peterhayes]

Gskid

The quick and easy answer is NO ...

I agree with that to a point BUT if you have a large or multi-monitors (mega TV) the more VRAM you have will preserve the performance.
On some routes scenarios my Dell 24" monitor at 1920 x 1200 resolution uses up to 1.8GB VRAM (all settings in TS2015 to the right and SGSS usage in NVI) - so if I had a 1GB card that would be throttling the output from my cpu, and I would see some slow down wrt frame rates/frame dwell times.
I now run a DTX 970 with 4GB VRAM on a 22" 3D monitor and TS2015 seems to be smoother and "faster" but that could be wishful thinking.
If you have a large monitor, TV or multi-monitors then VRAM does come into it to preserve frame rates/dwell times when displaying enormous amount of pixels.
You really need to match the cpu to the gpu otherwise you may get throttling of one or the other.
pH

[OldProf]

Thanks to GSkid for the link to the graphics card myths article and to Peter for his rebuttal based on large monitor usage (my current monitor measures 26" diagonally, so Peter's argument may well apply to me). Before, I was just confused about the various properties of graphics cards, now I'm overwhelmed, but better informed. It seems that I was probably best off with my GeForce GTX 550 Ti, but I'll give the GeForce GT 740 4GB GDDR5 that should arrive from dear old Amazon today a try before definitely deciding.

One good thing at least will come of all this: although I regularly blow the dust out of my computer, I learned the hard way that I'd been neglecting the graphics card's fan! There was a lot of fluff in there.

It's good to be part of a forum where knowledgeable members freely proffer informed advice.

[GSkid]

No problem! And it's not the monitor size, it's the resolution of the monitor that matters.

My 4GB GTX 850M has the slower DDR3 memory. There is a 2GB GTX 850M that uses the faster GDDR5 memory (it wasn't an option from HP), but the 4GB DDR3 version has it's GPU clocked higher than the 2GB GDDR5 model to help compensate. They pretty much perform the same.

More than likely the new card you have coming isn't going to improve things a ton, but that bigger 4GB might certainly help resolution and filtering-wise. What you should do is figure out the most you are willing to spend for a card. Then... start looking at cards in that price range and start inputting the model numbers of them into GPUBoss to find the model that scores the highest over your GeForce GTX 550 Ti overall. Check Amazon, Newegg, Tigerdirect, etc ...to find the best price. Look at reviews and customer feedback on models too when possible.

[thecanadianrail]

One good thing at least will come of all this: although I regularly blow the dust out of my computer, I learned the hard way that I'd been neglecting the graphics card's fan! There was a lot of fluff in there.

I learned this the hard way with my laptop, blew the cooling fan.... Still is broken and haven'tggotten around to fixing it, only reason my pc has not melted yet is because I've got a massive gaming laptop cooling fan underneath it.

[peterhayes]

Old Prof
Just to add to your confusion the GTX750, 970, 980 are the latest video cards from NVidia using the "MAXWELL" platform.
Which "supposedly" gives better performance in games??
The 750 is a 2GB card in some variants and does not need a separate power supply, for me it has the same performance as say a GTX670 and TS2015 runs quite well.
These new cards no longer support CSAA but have their own variant (MXAA/MSAA??) and that certainly seems to give a clearer picture wrt TS2015.
With the latest drivers from NVidia you can now downsample ie (DSR) Dynamic Super Resolution which can enhance any game that supports resolutions above 1920x1080 (eg TS2015).

What does DSR do? Simply put, it renders a game at a higher, more detailed resolution and intelligently shrinks the result back down to the resolution of your monitor, giving you 4K, 3840x2160-quality graphics on any screen.

Hence if you use DSR to get a better resolution in TS2015 then again the amount of VRAM on the card becomes very important. The GTX 970 with 4GB ram can display 4K resolution on my rig - but due to artefact which made the mouse cursor huge - I now use a DSR of 2K and that looks quite good.
More confusion
pH