1 - 11 of 11 Posts

#### brettjv

·
Joined
·
14,041 Posts
Discussion Starter · ·
Today I shall put the above-oft stated assertion to the test.

We hear manifestations of this claim in all manner of forms:

"Of course the 7970 smokes the 680 at high resolutions. It's the bandwidth!!!"
"660ti is a nice card, but it takes a dump at high resolutions. Look at that crappy 192-bit bus!"
"Of course the Titan isn't that great at 1080p. Have to crank up the resolution to take advantage of that 384-bit bus!"

So today, I've decided to put this pervasive theory to the test with my 670, in an attempt to see a) if it holds water, and b) if so, how much of the proverbial water does it hold

I'll be using the Valley benchmark, one that is known to be highly memory-bandwidth dependent. I'll be running it with all settings maxed, at two resolutions, the latter of which being precisely twice the size of the former, to wit: 1280x800, and 2560x1600.

At each resolution, I'll be running the test at stock memory speed (1502MHz), and again with a hefty overclock (1760, or (1760-1502)/1502 = 17.17%). We shall then assess the scaling (% fps increase/% bandwidth increase) provided by the additional bandwidth at each of the two resolutions. Granted, this is not the MOST optimal way of doing this sort of test, because increasing clock speed does not provide a 100% equal in every possible scenario equivalent to increasing the bus width, however ... a vast majority of the time ... it accomplishes the exact same thing.

Without further ADO, here are the results of said tests:
1280 x 800, 1502MHz: 75.0fps
1280 x 800, 1760MHz: 79.8fps

2560 x 1600, 1502MHz: 24.6fps
2560 x 1600, 1760MHz: 27.1fps

Running these math's through my high-tech supercomputer, I derive the following Bandwidth Scaling numbers:
@1280 resolution: (79.8-75.0)/75.0*100 = 6.4% delta fps / 17.7% delta bandwidth = 36.1% memory bandwidth scaling.
@2560 resolution: (27.1-24.6)/24.6*100 = 10.16% delta fps / 17.7% delta bandwidth = 57.4% memory bandwidth scaling.

So there you have it ... proof of a fairly positive nature that it is in fact true that as resolution increases, the relative usefulness of increased memory bandwidth also increases. Although the effect of doubling resolution does not quite result in a doubling of the bandwidth scaling, it's actually surprisingly close.

I would also add that number like 57% bandwidth scaling is incredibly high relative to most cards, and pretty damn solid evidence that the Kepler offerings came out with less memory bandwidth than they probably really should have. And if you AREN'T overclocking your memory because you've always heard it 'doesn't do much', well ... with Kepler ... it does A LOT.

I invite others to indulge in similar testing using other benchmarks or games, to see whether this is an across the board phenomenon, or what

·
##### Registered
Joined
·
410 Posts
I could tell you that my 660 gtx takes a dump at 1080p resolution !
Which just goes to show how awful these cards are designed. 660 has a lot less cuda cores then 660ti, so it's safe to assume that it will need less memory bandwith then its bigger brother 660ti.
And I am only playing at 1080p...

It seems to me there is more then 50% memory scaling on only 660gtx not ti and 1080p resolution !

I cant imagine how cripled 660 ti is...what was nvidia thinking ???

#### Vonnis

·
##### Registered
Joined
·
1,755 Posts
I did notice I get much more prominent gains from a memory overclock on my 680s than I did on my 480. Silly nVidia.

Quote:

I cant imagine how cripled 660 ti is...what was nvidia thinking ???
Something along the lines of
\$\$\$
, I imagine.

#### brettjv

·
Joined
·
14,041 Posts
Discussion Starter · ·
Quote:

I could tell you that my 660 gtx takes a dump at 1080p resolution !
Which just goes to show how awful these cards are designed. 660 has a lot less cuda cores then 660ti, so it's safe to assume that it will need less memory bandwith then its bigger brother 660ti.
And I am only playing at 1080p...

It seems to me there is more then 50% memory scaling on only 660gtx not ti and 1080p resolution !

I cant imagine how cripled 660 ti is...what was nvidia thinking ???
Well, in the defense of a card like the 660ti, it's really designed for 1080p, it does allow for pretty nice memory OC's in most cases which does help a lot, AND ... these Unigine tests are among the most memory-bandwidth dependent tests I've ever come across. I'd venture to say these bandwidth scaling numbers are just about the highest they could possibly be. This is likely a worst-case scenario IOW.

Not to mention, they give ya a full GTX670 core on the thing ... they had to cripple SOMEthing in order to sell it for less, otherwise 670/680 owners would've been really PO'd

Like I said in the OP, I'd be stoked if someone could run similar tests (and actually work out the calculations the way I did them) with some actual games. To do the test properly, you have to do at least four runs, two different resolutions, and two different memory clocks. We're not just looking to see the bandwidth scaling at one resolution, but rather, how scaling increases as resolution increases

By the way, your memory scaling is: ((32.6-30.8)/30.8) / ((3305-3005)/3005) = (.058 / .099) *100 = 58%

Basically the same as my memory scaling at 2560x1600.

·
##### Registered
Joined
·
410 Posts
Quote:
Originally Posted by brettjv

I'd venture to say these bandwidth scaling numbers are just about the highest they could possibly be. This is likely a worst-case scenario IOW.
Not really, I just tested Far Cry 3 and its even worse..forgive me I am too lazy to do the math it seems to be about 60% or even higher memory scaling

Quote:
Originally Posted by brettjv

Like I said in the OP, I'd be stoked if someone could run similar tests (and actually work out the calculations the way I did them) with some actual games. To do the test properly, you have to do at least four runs, two different resolutions, and two different memory clocks. We're not just looking to see the bandwidth scaling at one resolution, but rather, how scaling increases as resolution increases
Well I only have 1080p monitor, I am not even sure if I can run higher resolution, and I dont see the point in running lower then 1080p.
But it still has relavance, if a mere 660 gtx is clearly bandwith depraved at only 1080p, it pains clear picture what will happen at higher resolutions espacially with cards like 660 ti

#### brettjv

·
Joined
·
14,041 Posts
Discussion Starter · ·
Quote:

Well I only have 1080p monitor, I am not even sure if I can run higher resolution, and I dont see the point in running lower then 1080p.
Sure you can. My monitor is 1920x1200. I used downsampling to test at 2560x1600

And the 'point' is just to test, to see how increasing resolution increases bandwidth scaling. Increasing our collective knowledge, as it were. FC3 doesn't lend itself to easy testing though, since there's no benchmark for it.

·
##### Registered
Joined
·
410 Posts
Ok, I re-run the tests I already done on 1440p.

It doesnt seem higher resolution increase memory scaling much over 1080p, if at all, atleast on 660 gtx.
In case of Far Cry 3, memory scaling is actually reduced on 1440p compared to 1080p.
And dont think I didnt test Far Cry 3 properly just because there is no benchmark for it, just loading the game at the same exact spot couople of times on same clocks to confirm there is no any fps variations should do the trick. ( its always 31 fps on 3300 memory and always 30 fps on 3000 memory not metter how many times I load )
Maybe thats because its more shader limited at 1440p then its at 1080p ?

#### Shiftstealth

·
##### Network Enthusiast
Joined
·
4,793 Posts
WHY CAN I NO REP?1?!

Thanks <3

#### Scorpion49

·
Joined
·
8,370 Posts
Contribootin SLI 660 numbers:

Disclaimer - I had to use 4x AA because my cards were vram limited at 2560x1600 with 8x and FPS dropped to like 2.

1280x800 @ 1202/1400

1280x800 @ 1202/1750

2560x1600 @ 1202/1400

2560x1600 @ 1202/1750

1 - 11 of 11 Posts