• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

800 FSB is here thugsrook format

M

mroleg

Senior member
P4 3.0 Ghz 800FSB/Asus P4C800-D/2x512 OCZ 3500EL/80Gig WD SE/GF FX 5800 Ultra/ALPHA PAL8942 w. TT Smart FAN
System Performance - Turbo
VCORE - DEFAULT
DDR Voltage - 2.75

RAM freq - 460 Mhz
RAM settings 2.5-3-3-8
Sandra scores:
FSB-buff-unbuff
230-5300-4437
3Dmark03 Score: 5891

Prime95 - 6hrs - rock solid.


RAM freq - 466 Mhz
RAM settings 2.5-3-3-8
Sandra scores:
FSB-buff-unbuff
233-5326-4475
Prime95 - 6hrs - rock solid.

RAM freq - 470 Mhz
RAM settings 2.5-3-2-7 2.85v
Sandra scores:
FSB-buff-unbuff
235-5407-4535
------------------------------------------------------------------
3dmark03 score: 6250
------------------------------------------------------------------
PCmark02 score:
CPU-memory-hdd
8704-10141-1091
------------------------------------------------------------------
Aida Score:
MemWrite-1791 MemRead-5162
--------------------------------------------------------------------
Comanche 4 Benchmark Results
1024x768x32 FSAA=0 VSync=OFF DXTC=ENABLED AUDIO=OFF SHADERS=ENABLED
Frames per second: 70.74 avg
Tris per second: 14,088,386 avg
----------------------------------------------------------------------------
PiFast
57.92 Seconds
-------------------------------------------------------------------------------
Prime95 - 6hrs - rock solid.
Temps:
32-Idle
38-Loaded
These are pretty accurate because Asusprobe and Nexus temps are insync. (RARE????)
******anything higher is not stable********


RAM freq - 392.1 Mhz
RAM settings 2.5-3-3-8
Sandra scores:
FSB-buff-unbuff
245-5152-3910
 
Originally posted by: snowwie
hmmm...thugs was right
more mem bandwidth is better
what a suprise

why didn't you lower the timings @ 245fsb?
Click to expand...
Sanrda mem scores are hardly a good test of system performance.
rolleye.gif
The format is a bit hard to follow since CPU speed is not included

From what I see here, he has

230 MHz FSB
3.45 GHz
DDR460

233 MHz FSB
3.49 GHz
DDR466

245 FSB
3.67 GHz
DDR392 (5:4)

I would say on this rig, the extra 180 Mhz CPU speed should edge out the loss of 74 MHz of DDR speed.

And yes, the DDR392 setting should be tested @ 2-2-2-6 to get valid results

Another thing is, the 3.0 is not the ideal candidate for this type of testing. The 2.4C, 2.6C, and to a lesser degree, the 2.8C will hit some very high FSB speeds if they turn out to be good overclockers. Using the 3.67 GHz example from above,

2.4C @ 3.67 = 305 MHz FSB
2.6C @ 3.67 = 282 MHz FSB
2.8C @ 3.67 = 262 MHz FSB

A setup like that will need mem dividers, no doubt.
 
well, I see 466MHz ram and 233fsb is faster than 392MHz ram and 245fsb

while that is a significant drop in ddr speed (74MHz, or 37?), I think that +12fsb is more significant, and don't forget quad-pumped
man, I really wish there were smaller ratios so mem bandwidth isn't hurt so much when we overclock

I just remember when we were talking about overclocking too high for ddr to be synchronious, everyone was saying that we should ignore the mem bandwidth for the sake of a higher fsb and CPU speed, and thugsrook was arguing for the side of mem bandwidth

mroleg inched his fsb up only so much as to not forsake the mem bandwidth, then he dropped the ratio and pushed the cpu all the way, and it looks like backing up the cpu and keeping the memory synchronious is better.
 
Originally posted by: snowwie
well, I see 466MHz ram and 233fsb is faster than 392MHz ram and 245fsb

while that is a significant drop in ddr speed (74MHz, or 37?), I think that +12fsb is more significant, and don't forget quad-pumped
man, I really wish there were smaller ratios so mem bandwidth isn't hurt so much when we overclock

I just remember when we were talking about overclocking too high for ddr to be synchronious, everyone was saying that we should ignore the mem bandwidth for the sake of a higher fsb and CPU speed, and thugsrook was arguing for the side of mem bandwidth

mroleg inched his fsb up only so much as to not forsake the mem bandwidth, then he dropped the ratio and pushed the cpu all the way, and it looks like backing up the cpu and keeping the memory synchronious is better.
Click to expand...
Well, i guess if you use your PC for nothing other than Sandra mem benches, yes that is true. If you actually use it for real applications, its not.

Again, Sandra mem benches mean nothing. Haven't you ever looked at this before?
 
yes but the actual mem bandwidth (sandra) you get makes a really big difference in performance

there are many (true) arguements that faster ddr doesn't mean faster ram (sandra)

many people have said that if you change your mem ratio from 1:1 to 3:4 (assuming your ram can handle it), and keep all other sttings the same, your mem bandwidth (sandra) may not necessarily be better

edit:

okay, instead of using sandra as an example, think of Unreal scores, or encoding/decoding

they're all the same, telling you how fast your computer is
 
we need some 3dmark benchies before drawing a conclusion.

obviously 3.67g is faster then 3.4g ~ but is the system faster?
that 3.67g could be easily choked if 392ddr aint enough bandwidth.

or more to the point ~ do we need 200fsb systems for this at all?
how would a 1:1 ratio 533 chip compare @ 3.67g?
 
yes but the actual mem bandwidth (sandra) you get makes a really big difference in performance
Click to expand...
Sorry, but no, it doesn't. I cant tell you how many benches I've seen with a 20 - 30% Sandra mem increase and a 2 - 5 % real world increase. Sandra mem score is the most lopsided out there.
many people have said that if you change your mem ratio from 1:1 to 3:4 (assuming your ram can handle it), and keep all other sttings the same, your mem bandwidth (sandra) may not necessarily be better
Click to expand...
Who said that? Not me or anyone else I've ever seen here. As a matter of fact, I was one of the first ones here running 3:4 back in the 845D days telling people about the benefit.
okay, instead of using sandra as an example, think of Unreal scores, or encoding/decoding

they're all the same, telling you how fast your computer is
Click to expand...
Now you're talking! Real world benches are what we need here!
You should read this article
and this one too

mroleg, didn't mean to hijack your thread (sorry)😱 Could you do a few more benches such as games, video encoding, mp3 encoding or whatnot? Also please test DDR392 @ tighter timings such as 2-2-2-6 or even 2-2-2-5. That should not be a problem at that DDR speed.

Thanks!
 
Originally posted by: THUGSROOK
we need some 3dmark benchies before drawing a conclusion.

obviously 3.67g is faster then 3.4g ~ but is the system faster?
that 3.67g could be easily choked if 392ddr aint enough bandwidth.

or more to the point ~ do we need 200fsb systems for this at all?
how would a 1:1 ratio 533 chip compare @ 3.67g?
Click to expand...
Good points. What I like about the 800 FSB CPU's is they have HT, which will give a nice boost for "free". Also, as your P4m testing showed, the high FSB speeds really perk things up. Again, the 3.0 chips is not really the ideal CPU for doing this. The 2.4, 2.6, 2.8C CPUs have the potential to hit some mad FSB speeds!

 
sorry, I didn't mean sandra specifically, I meant benchmarks in general

and that overclockers article is a perfect example of someone telling us to "screw mem bandwidth, raw CPU speed is all that matters"

I was thinking of that article when I was writing my posts

Just look at some benchmarks of the latest chipsets (845PE, SiS655, granitebay, canterwood, nForce2, KT400, etc.), and some show DDR333 as being faster than DDR400
I'll try to remember where I saw specific examples of faster memory hurting performance compared to slower

edit:

but you guys are right.
all we have right now are sandra mem scores
we need more benches to really know which is faster
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top