Does decreasing the multi for higher FSB improve or decrease frame rates in games?

[Solaris17]

now would oc'ing with a lower multi require less voltage on cpu?

[HTC]

Originally Posted by EastCoasthandle

No it didn't from what I observed. Except for when I used a PL6 instead of PL7 but I provided both. If you look at my results the timing didn't change.

Right: the only thing i noticed was the change in the NB strap.

Originally Posted by Solaris17

now would oc'ing with a lower multi require less voltage on cpu?

1 increment only, in the VCore BIOS setting. It's only a tad bit lower, but it was lower.

[niko084]

Originally Posted by HTC

Right: the only thing i noticed was the change in the NB strap.



1 increment only, in the VCore BIOS setting. It's only a tad bit lower, but it was lower.

I got the same thing when I dropped to 7x @ 3.2, I dropped the voltage on my chip by 1 increment also and remained stable, although my ram took .05 more.

[HTC]

Originally Posted by niko084

I got the same thing when I dropped to 7x @ 3.2, I dropped the voltage on my chip by 1 increment also and remained stable, although my ram took .05 more.

The BIG problem with raising the FSB is the higher PL (tRD) value:

http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=9

The bigger the PL, the slower your RAM gets.


@ EastCoasthandle: could you try, using the settings of your previous pics (including the one with PL6), to RAR about 1 GB of files and post your results?

There will be differences ...


Also, a dumb question:

Don't games use graphic card memory only except during load times?

[niko084]

Games use both heavily, more heavily depending on the game.

[HTC]

Originally Posted by niko084

Games use both heavily, more heavily depending on the game.

But isn't the graphic card's RAM faster the the system one? If so, a little increase in system RAM speed isn't going to make much difference ...

It's like system RAM VS pagefile!

[niko084]

Originally Posted by HTC

But isn't the graphic card's RAM faster the the system one? If so, a little increase in system RAM speed isn't going to make much difference ...

It's like system RAM VS pagefile!

Video ram is only used for graphics. But yes video ram is generally faster.

[HTC]

And it would explain why there's little to no difference in games.

[EastCoasthandle]

Originally Posted by HTC

@ EastCoasthandle: could you try, using the settings of your previous pics (including the one with PL6), to RAR about 1 GB of files and post your results?

There will be differences ...


Also, a dumb question:

Don't games use graphic card memory only except during load times?

This is not 1gig of files but it does prove the point.




I will have to do the others at a later time/date...sorry



400x9 PL6 = 1793





400x9 PL7 = 1755

[EastCoasthandle]

450x8 @ 1080 = 1812 using


450x8 @ 1128 = 1857

[HTC]

That's about 7% difference (in the time), which is significant. Maybe 1 GB of files isn't enough for a comparison

[niko084]

Well I ran it-

Supreme commander e6750 @ 2.66 DDR2-666 42.1 FPS Average
Supreme commander e6750 @ 2.66 DDR2-886 45.8 FPS Average

So as I said before ram intensive game with a good difference in speed shows a change.
By the way those averages were a average of 10 runs, and at 886 it was always 2-4 fps ahead.

[Mussels]

Originally Posted by niko084

Well I ran it-

Supreme commander e6750 @ 2.66 DDR2-666 42.1 FPS Average
Supreme commander e6750 @ 2.66 DDR2-886 45.8 FPS Average

So as I said before ram intensive game with a good difference in speed shows a change.
By the way those averages were a average of 10 runs, and at 886 it was always 2-4 fps ahead.

what was the sim speed of the game. sup coms FPS is very erratic, especially depending on zoom level, while the sim speed is a good indication of ram and CPU performance.

[niko084]

Originally Posted by Mussels

what was the sim speed of the game. sup coms FPS is very erratic, especially depending on zoom level, while the sim speed is a good indication of ram and CPU performance.

The actual benchmark the game has.

[Mussels]

Originally Posted by niko084

The actual benchmark the game has.

the in game benchmark was reported on their games forums to be a bit dodgy lately... but ok, thats a fair test for FPS, especially since you ran it multiple times.

[EastCoasthandle]

Originally Posted by niko084

Well I ran it-

Supreme commander e6750 @ 2.66 DDR2-666 42.1 FPS Average
Supreme commander e6750 @ 2.66 DDR2-886 45.8 FPS Average

So as I said before ram intensive game with a good difference in speed shows a change.
By the way those averages were a average of 10 runs, and at 886 it was always 2-4 fps ahead.

I have to agree, the benchmark results for this game are erratic, regardless of the number of times you repeat them. Using a game that offers consistent benchmarks from one run to the next would be a bit more creditable in my book. Here is another example:
1428FSB = 17248
1776FSB = 17493
1800FSB = 17447

It's also interesting that when they tested the PC2-9136 with a 780i board from one FSB to another (using EPP) game results are nearly the same.

[Mussels]

supreme commander is only worth testing if you use the perftest results, but thats DEFINATELY broken in the latest patches/expansion (its capped and hardly ever changes results on most systems)

The best way to use that is to run a game, and save a replay - go back to the same time and see what the results are. (for example, if you get -3 sim speed exactly one hour in, the replay would be an accurate representation of that same CPU/ram load, and any improvements would show up there)

Edit: heres a screenshot. Esp in multiplayer games, the game slows down to match the slowest player - so even tho the game may be set to +10 speed, you'd be getting 90FPS because some other dude has crippled you down to -2.

As shown in this screeny, even tho there is only a totally wasted AI left, i'm still stuck at +4 since the AI in this game... well it kinda cripples the speed.

[niko084]

I don't have the expansion, and I just reinstalled it to run the test, no patches at all.
I'll run some Crysis tests or Vegas:2 tests later here.

[EastCoasthandle]

400x9 @ 1066
Min 26.33 FPS
Max 46.65 FPS
Avg 38.735 FPS



450x8 @ 1080
Min 26.87 FPS
Max 46.68 FPS
Avg 38.775



450x8 @ 1128
Min 27.79 FPS
Max 46.16 FPS
Avg 38.785
Sorry, I forgot to scroll the doc up but the score is found on the benchmark tool itself.


Minimal frame rates results did increase but minuscule in my book. Max frame rates were sporadic at best and didnt' show linear gains from one dram frequency to the next. I didn't include the CPU/Dram settings in all of the photos this time around. Sorry about that!

[EastCoasthandle]

I wonder what games people are playing to say they notice a difference?

[eidairaman1]

I did run some benchmarks, i found out the FSB being at 207 and multi at 11 was the same as the FSB being at 222x10.5, so its best off to have a Slightly higher Multi.

[HTC]

Currently, i know of no games that increase by a considerable amount (more then 1 FPS or so), which is why i voted for only memory bench progs.

The same can't be said for programs that make heavy use of memory, such as encoding ones, though:


This is with a PL of 6


This is with a PL of 7

The above tests were conducted on the same movie: i only changed the PL, the multi and FSB (and some voltages: different ones in one or two instances).

Had i used a PL of 7 for both, the disparity should be even bigger, i believe.

EDIT

That's slightly over 6.5% reduction: from 757 seconds to 707 seconds!

It's a significant number!!

[niko084]

Originally Posted by EastCoasthandle

I wonder what games people are playing to say they notice a difference?

Their ram is probably a good bit slower, most of us with faster machines are starting our ram at 800+ our ram is not a bottleneck.

Increasing the FSB and dropping your multi is really only going to help you in a ram bottleneck type issue.

[EastCoasthandle]

PL7
9x multi
1093 Dram Frequency

All settings are the same as before.