Gaming PC Upgrade

Theocnoob said:

I disagree. I say you spend the extra $400 on SLI 780s. You can't brag to your friends about 970s. JM2C.

Click to expand...

If I was a teenager, I would care about that reasoning.

I look at it like Ninja does... using his example, 980 SLI costs almost 60% more than SLI 970's and yields, what, a 15% FPS increase? That math doesn't add up in my world. Now, will it ever be 1:1... usually not, but that is a HUGE difference.

EarthDog said:

If I was a teenager, I would care about that reasoning.

I look at it like Ninja does... using his example, 980 SLI costs almost 60% more than SLI 970's and yields, what, a 15% FPS increase? That math doesn't add up in my world. Now, will it ever be 1:1... usually not, but that is a HUGE difference.

Click to expand...

Teenager? Come on dude.. Most of us have high end PCs because we wanted them, not because we needed them. No home office setup needs a GTX 970 and an overclocked i5. I say, if you're already throwing $1500 at a given situation, why not add another $200 and get the best card you can? It's never been a wise financial decision to get the top tier card. Ever. It's for E-Peen. Pure E-Peen.

^Agreed.

How much would the skylake stuff cost when its released? would it be about the same as the devils canyon stuff at the moment?

I'd imagine so.. only time will tell though.

he could brag to me, I run 2 gtx760's in sli!!!!!

REALY, fast ram to game??? dont bother. in fs 10, I have run, 1600, 1866, 2000, 2133 and 2400 ram, guess what, 2-5 fps, ram means nothing in game.
My 760's do all i need them to do, i can afford the 780ti or a new 980 or two, I game at 1080p with sync on so 2 gigs of ram are enough, for whats coming down the pike game wise, 3 gigs should hold you over with a single 970, now if your going multi monitor at higher res, that changes the story.
a single 970 for the card and 8 gigs of 1866 ram will be more than impressive.

Caddi- you think 5FPS is nothing? Come on, dude. 5FPS is a lot.
Also, considering the price difference from 1600 to 2133 you'd be nuts not to get 2133. JM2C.

Again with the flat fps gains. 5fps is huge at 25fps but it means nothing at all at 250fps.

http://www.anandtech.com/show/7364/memory-scaling-on-haswell/7

Tl:dr with a single card 1333 cas 9 basically ties 2400 cas 9. With trifire, gains are anywhere from nothing to 20% depending on game. And again, this is at lower resolution specifically to eliminate gpu bottlenecks and hilight memory performance. With real world resolutions the gpu will be the bottleneck.


As far as "feel" goes, o once pulled a stick of 1866 ram out of my system to test a motherboard for my brother. Leaving me in single channel mode. I forgot I took it out and gamed for like a week without ever noticing any performance hit. Half of my bandwidth gone and I couldn't even tell.

^^I have a hard time believing you noticed no diff between single and dual chanel. Did you run a frame counter before and after and run the same benchmark?

Of course I didn't. And I'm not saying single channel isn't slower. I'm just saying it was such a small performance hit I forgot I had a stick out. I don't ever have frame counters running unless I'm testing something. I just play my games and if they are smooth then I'm good.

But now you have me curious how big or small of an impact it will make. If I remember when I get home from work I'll run a couple of benchmarks.

Edit: Just a quick update, this isnt an extensive test by any means, but the tomb raider canned benchmark at 2560x1440 ultimate w/fxaa hits 106 average with dual channel ram and 105 average with single channel. Thats 1866 9-10-10 ddr3.

Firestrike with dual channel mem was 15644 overall, 8763 physics, and 23712 graphics. Firestrike with single channel mem was 15303 overall, 8684 physics, 23064 graphics.

a single run of each, so there was no accounting for the small amout of variance run to run, but its pretty clear that at leas tthose two tests arent really limited by memory bandwidth.