Highfsb Vs 2,2,2,5

[Mtemel]

hyperthectically speaking if .......


u r running 260 fsb 1t 1:1 with nice timings of 2,2,2,10

my question is this

lets jus say we are using 3dmarks01 and score 25000 with this setup?

then i change my timings to 2.5,3,3,7
and run 3dmardks01 again wot difference in score would i get from these 2 diff timings?

also do the gains get more the faster i run my cpu htt and mem??

eg if i had 600 marks increase with tight timings @ 200 htt 1t 1:1 etc would my gains become more if i raised the cpu htt and mem to 270 htt 1t 1:1

or would it still be the same 600 score gains at these higher settings??

i hope u understand what i am getting at

[reject]

i dont know in marks, but maybe 20 fps in certain tests, such as CH, LH
tight timings on a64 beat looser timings no matter what the fsb
except in superpi, that likes bandwidth. but it would have to be 2-2-2 at 200 vs 2-3-3 at 260 or so. timings is the key
and the faster the cpu the less it relies on fsb and more on timings

[bacon]

me and philonslaught did a comparison with his tccd rev2 at 280 (2.5-3-3-7) and my VX at 255 and 260 2-2-2-5), neither had been tweaked with A64 tweaker

He scored about 100mbsec more with 20mhz or so and amrginally less cpu speed

http://forums.overclockers.co.uk/showthread.php?s=&threadid=17332159

thats only sandra not 3dmark tho, so make what you will out of our comparison

[blinky]

if and when i get tccd, ill do a direct compare to see the score differences of bh5 with 2-2-2-5 and tccd with 2.5-3-3-7, im guessing its about 30mhz that ud have to gain with tccd to make up for the latencies

[bacon]

its in my comparison id guess its closer to 20mhz (may even be less) to get the gain back, tho iv still yet to tighten those timings

[Mtemel]

cool thx lads looking forward to ur results blinky :d

btw bacon wots the max fsb u can get out of ur vx mem on 2.5,3,3,6 timings?

[mikead_99]

The 38.xk score is with 270x12, 2x512 TCCD @ 2.5-3-3-10, the 39.x score is only 250x13, 2x256 BH-5 @ 2-2-2-5. Unfortunately even though the BH-5 will do 270x10 w/2-2-2, I can't get stable 270 2-2-2 at the cpu's max clocks, but I think the gains would be awesome. 2-2-2 with even a decent speed is where it's at. It looks like car low was the biggest winner, and I know we can say I didn't get a good run at 270 with TCCD, but both are max'ed bugged runs, 2-2-2 is just that much better for that test.

http://mysite.verizon.net/res6otqr/sitebuildercontent/sitebuilderpictures/222_2.533.jpg

[bacon]

cool thx lads looking forward to ur results blinky :d

btw bacon wots the max fsb u can get out of ur vx mem on 2.5,3,3,6 timings?


will try tonight if i have time :)

[Malachia]

then the question is this now, 2*256 single channel, will be a lot better than 1*256 single channel(ovviously....)

[Rabbi_NZ]

with my OCZ VX....

3Dmark01
10x250 2-2-2-0 : 22449
10x250 2.5-3-3-10: 22290

SuperPI 1MB
10x250 2-2-2-0 : 34s
10x250 2.5-3-3-10: 35s

SiSoft Sandra Memory Bandwidth
10x250 2-2-2-0 : 3879/3882
10x250 2.5-3-3-10: 3868/3871

[Geforce4ti4200]

3Dmark01
10x250 2-2-2-0 : 22449
10x250 2.5-3-3-10: 22290


That is only 159 marks gained. I gained exactly 158 marks when I went bh5 from corsair value back on my winchester, err s754 newcastle. I guess I have too much winchester on my mind. timings are worth 30fsb which in turn is worth 90MHz cpu clocks. If someone is getting like 500 marks it means he got a very, very bad run with the slower ram and a very, very good run with the better ram. I really only gained in both lobbies with bh5, no other test went up.

[reject]

im AM3 it can be up to 5k points difference

[Rabbi_NZ]

im AM3 it can be up to 5k points difference
will post results in a day or so

[Zeus]

http://www.gskill.com/forum/index.php?showtopic=74

Here you'll get some idea how 2,5-3-3-x does vs 2-2-2-x in 3D mark.

Enjoy. ;)

[Mtemel]

thx alot guys from this info i think i need the ram that ll do 300 htt @ 2.5,3,3,7 rather than the low lat vx stuff.

again thx alot for the help :D

[mikead_99]

If someone is getting like 500 marks it means he got a very, very bad run with the slower ram and a very, very good run with the better ram. I really only gained in both lobbies with bh5, no other test went up.

Yeah, I think the compare I showed isn't the most valid. With the TCCD run, I didn't test out CodeRed's Tweaker at that time, the RAM was new, and I only tightened a few settings, and I know now that twtr was set too tight. This probably induced some errors, lowering the score. The BH-5 I have I know like the back of my hand, so it was pretty darn well optimized. Now that I know the tweaks my TCCD can handle, I will probably go ahead and run a 250x11 compare one time for my own gratification, both sets of mem as tweaked as they can handle well.

Regarding AM3 differences, I certainly didn't gain 5k. More like 1200 points for me. Still not a bad gain though.