I have upgraded my EVGA GTX 670 FTW to the MSI GTX 970 Gaming OC. Kludger has a similar system to me and asked if I could let him know how the upgrade goes and suggested that I share my findings on the forum.

I have benched games that you don't often see in graphics card reviews just to give a taste of how the upgrade affects the types of games we play. I cranked up all settings on the games and ran built in game benchmarks where available.

My system is as follows:

ASUS P8P67, Intel i5 2500K, 8GB Corsair vengeance DDR3 1600MHz RAM, Samsung 840 EVO 500GB & 250GB, various hard drives etc etc.

As you can see, it has highlighted that the i5 2500k has become a bottleneck for sims and CPU intensive games.

Also, not on the charts is the Rome 2 Benchmark which got 38 fps with the GTX 670 and 66.4 fps with the GTX 970.

The civilization is the new game but couldn't remember the title hehe i just call it Civ future

I hope people find these useful, but please don't use this as excuse to start waiving e-peens it's just me trying to help out a friend and thought I'd share.

Min, Average, max fps


Average fps

Thanks for sharing Tigerwulf! A GTX9x0 is on the list for my next upgrade when Broadwell-K shows up. What resolution do you run at?

/KC

Thanks Keycat! I run games in 1080p. I do have triple monitors but rarely use that res because of the edge of screen stretching. Thinking of upgrading to 1440p though.

I should have done a couple of triple screen tests but I already had the card for two nights without playin a game with it and just wanted to get the benches done so I coild play hehe

Thanks Tigerwulf. The only game I care about is FSX. From your comparison test, getting a significantly more powerful GPU has almost no measurable affect on frame rates. Can I assume this means FSX is totally CPU bound? To get better fps in FSX, then, I will have to significantly upgrade my CPU instead of my GPU? I'll be ta GTX 580 (my current GPU) plugged into your PC would get the same fps as your 670 and 970. That's depressing.

Yup it would appear so Plainsman. It's something we've always known to some extent but now that I have finally done an actual benchmarked test using the PMDG approach flight I can see that the GPU change had no real impact on framerate. It does appear that we need much faster CPUs.

I was surprised by the ARMA3 result though and just ran a quick test again whilst monitoring the CPU and GPU and on Ultra settings with vsync off which I used for the benchmark the CPU was only at 50%-60% and the GPU at 40% and the framerate was limited to:

Avg: 39.183 - Min: 30 - Max: 54

I changed the setting to low and the framerate was:

Avg: 67.383 - Min: 59 - Max: 80 GPU usage was 25% and CPU usage about 80%. It leads me to believe that ARMA3 is bugged somehow in the coding and not using the hardware to it's full potential and may be a misleading indicator for hardware performance.

THanks for posting your results, Tigerwulf. I have a similar system which I intend on keeping a while longer, but my next upgrade will likely be to an NVidia GPU (in conjunction with a GSYNC monitor).

One way to definitively confirm the CPU-bound nature of FSX is to overclock (or downclock) your i2500k and see if the change in fps is commensurate. Just as with ARMA3, there may be a coding limit... come to think of it, did you remember to change the "target fps" (I think that's what it's called) in the graphic settings in FSX? It looks as if you had them locked in at 45fps.
C_G

Yup I set the frame limit on all games to unlimited and made sure I turned off Vsync in game and also in any Nvidia panels etc.

I will downclock my i5 2500k (already have it at 4.5Ghz) and see how it affects FSX.

I'm anxious to hear the results.

You might want to think long and hard about putting more cash into a CPU Plainsman, I am now.

Thanks Tigerwulf, that looks like a great GPU upgrade for the games/sims that take advantage of the GPU, and as usual FSX remains CPU bound

Thanks for sharing those results, I'll be looking at this same upgrade this next year for my identical system.

Are you sure about Arma 3?

Tom's uses it in their gpu reviews, the 970 has 63 min and 70 avg, the 770 which is a generation later than your older card had 49 min and 58 avg.

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-980-970-maxwell,3941-9.html

Thanks Tigerwulf. The video card doesn't even make a friggin' difference. It's all CPU. I gotta have my FSX.

Sorry Remon I can only go by what Fraps told me on the benchmark I made. I don't know how Tomshardware made their benchmark so can't comment on the difference.

Definitely not enough to warrant an upgrade for FSX Plainsman

Thanks Tigerwulf.
Interesting that FSX is so CPU-bound. On the other hand you are getting decent frames.
Personally, I find FSX perfectly flyable so long as FPS stay above 20fps at low alt above a city. So, for me, this doesn't suggest that I'd need to upgrade the mono/CPU, just that my FSX experience wouldn't change much if only the GPU is upgraded. thanks again!

I couldn't find what Tomshardware use for their benchmark, Remon, but I have found another site that have benchmarked GPUs using ARMA3.

Here - http://www.techspot.com/review/712-arma-3-benchmarks/

I'm obviously not going to reinstall the GTX670 but I have run the infantry showcase as they do using Ultra with Vsync off and got the following results:

Frames: 4712 - Time: 90000ms - Avg: 52.356 - Min: 46 - Max: 63

Which is very good if you compare with their results of previous generations cards here:

http://www.techspot.com/review/712-arma-3-benchmarks/page4.html

The GTX 670 gets average 31 fps (although this is at 1920 x 1200 rather than 1920 x 1080)

It looks even better for the MSI GTX 970 when I overclock the card from it's effective base clock (using GPU-Z measuring clock whilst ARMA3 is running) of 1329 MHz and effective memory clock of 7GHz to 1469 MHz and 8GHz.

I got the following result from the same Infantry showcase run through:

Frames: 5075 - Time: 90000ms - Avg: 56.389 - Min: 50 - Max: 68

All I can surmise is that my benchmark uses some limiting factor of the game engine, probably AI, because my benchmark had me driving down a road in a hunter flanked by groups of infantry, some helos in the air, platoons of tanks etc.

Try lowering the gfx settings. If you get no significant gains in performance it should be the cpu.

I get massive gains on the infantry showcase and 'normal' play when lowering settings. I also get frames in the 50s when playing normally on ultra. Like I said it was the benchmark that I had created that had caused the issue with low frames. It must have been all the AI I had added.

It's really easy to make Arma's FPS tank if you put too much stuff in. The empty world runs very fast, static objects lower it a little, and active objects controlled by AI make a big dent.

Also, having Arma on an SSD instead of a standard HD helps average performance by getting rid of the I/O access hits as it streams the world around you.

I have a GTX 770 and an i7-2600k (still at stock speeds) and A3 gives me no real trouble aside from really big intense missions. However, I may start OC'ing my CPU now to see if I can get some more life out of it. If the OC makes it start to have trouble, then I'll upgrade it as it's getting pretty old now (I bought it right after release in early 2011). Actually I just realized my CPU is older than my younger daughter born April 2011.



The Jedi Master

I own an Intel Ivy Bridge 3570 that easily overclocks to 4.5 GHZ.
Don't bother, with Arma 3 the difference is negligible.
Been there done that Got the T-Shirt.

Good test Tigerwulf! I am also going to get the GTX970, as my good old GTX570 is sometimes a bit slow with IL2 BOS. I thought that my new I7 would chew the game for breakfast, but the 570 is slowing it down. I guess FSX is the only flightsim where a good videocard has abolutely no effect at all.