Real-world impressions of going from P4 to a faster P4...BENCHMARKS! ALSO...MUST remove 1 GB of memory to run stable 1:1

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

oldfart

Lifer
Dec 2, 1999
10,207
0
0
Originally posted by: fkloster
Memory is on the way back Old Fart.... still not sure which to get....somewhat confused.
You aren't alone. Memory for a P4 DDR rig is not an easy thing. The best thing is to start with your FSB speed and work back (if possible).

For you, you run a 266 MHz FSB speed. Is this the limit of your CPU, or is memory holding you back?

Assuming 266 is where you will stay, you can work the numbers from there.

266 MHz FSB 1:1 ratio = DDR532
266 MHz FSB 5:4 ratio = DDR426

To use DDR532, use need some very expensive ram, and you will also be stuck with loose mem timings.

DDR426, can use tighter timings. Testing has shown that the tight timings with the lower DDR speed is as fast or faster than the high speed DDR with the loose timings.

It is hard to find ram these days that will do tight timings. Ram mfgrs have tuned their mem for high DDR speeds and slow timings. The "bigger" number such as PC4000 sells even though it is not all it is cracked up to be.

The best deal I see out there right now is the Mushkin PC3500 L2 1 Gig (2 x 512) @ Newegg for $289. This was ~$350 a short time ago. This ram will run tight timings. How tight on your rig is hard to say for sure. It will do much better than 3-4-4-8 for sure. This ram was reviewed on the main page not too long ago and is highly rated.

Anandtech review

HTH
 

hurrikaane

Member
Oct 4, 2003
169
0
0
That mushkin runs at 2-3-2-6 at 216mhz. 5:4 ratio and a 2.4c gives ya 3200mhz and a 270 fsb. Not bad with those memory timings.
 

NFS4

No Lifer
Oct 9, 1999
72,636
47
91
Originally posted by: Dug
Now I have to buy some Raptors. My addiction for speed is killing me. (and my bank account)

BTW NFS4, why are you using a higher stripe size?

I thought that a higher stripe size meant greater performance?
 

Dug

Diamond Member
Jun 6, 2000
3,469
6
81
Originally posted by: NFS4
Originally posted by: Dug
Now I have to buy some Raptors. My addiction for speed is killing me. (and my bank account)

BTW NFS4, why are you using a higher stripe size?

I thought that a higher stripe size meant greater performance?

If you are transferring large files. For every day use, games, os, a smaller size is much better.
 

fkloster

Diamond Member
Dec 16, 1999
4,171
0
0
For you, you run a 266 MHz FSB speed. Is this the limit of your CPU, or is memory holding you back?

-OldFart

Ok... I can get to desktop @ 275fsb but can't seem to run nothin....
266fsb is stable...... but can't run even 1:1 with my 'supposed' 500mhz memory!!! (can't run @ 532 with loose timings)
250fsb is stable...... but can't run even 1:1 with my 'supposed' 500mhz memory... (can't run @ 500 with loose timings)
 

NFS4

No Lifer
Oct 9, 1999
72,636
47
91
Originally posted by: Dug
Originally posted by: NFS4
Originally posted by: Dug
Now I have to buy some Raptors. My addiction for speed is killing me. (and my bank account)

BTW NFS4, why are you using a higher stripe size?

I thought that a higher stripe size meant greater performance?

If you are transferring large files. For every day use, games, os, a smaller size is much better.

Is there like a "stripe guide" that explains all of this? Not that I don't trust you, it's just that I want to see some verification/test results...etc.
 

Dug

Diamond Member
Jun 6, 2000
3,469
6
81
Check this Thread for a quick blurb on someone going down in stripe size.
shuttleteam claims the 8k is the best after running benchmarks in all stripe sizes.

fkloster
send that damn memory back. If you can't run 1:1 at 250fsb then there is something wrong. It is definately not your cpu, becuase your stable at 5:4.

Maybe it has something to do with running 2G of the stuff. I don't remember seeing anyone benchmark with that much before. But you can verify that by taking out two sticks and testing.



 

fkloster

Diamond Member
Dec 16, 1999
4,171
0
0
Ok guys... I'm exhausted but here is the benchmarks....anything look about right?

6,650 3dMarks build 330

50,453 AquaMark3

Sandra Standard v. 2002.6.8.97

CPU Arithmetic Benchmark:

Dhrystone ALU = 7,421 Mips
Whetstone FPU = 6,496 Mflops

CPU Multimedia Benchmark:

Integer = 14,921 it/s
Floating Point = 22,662 it/s

Filesystem Benchmark :) :

Drive Index = 67,790 mb/s

Memory Bandwidth Benchmark:

Ram Int Buffered = 6,145 mb/s
Ram Float Buffered = 6,173 mb/s

Cache & Memory Benchmark:

2kb-----34,204 mb/s
4kb-----34,978 mb/s
8kb-----28,355 mb/s
16kb-----24,797 mb/s
32kb-----25,366 mb/s
64kb-----25,381 mb/s
128kb-----25,808 mb/s
256kb-----23,542 mb/s
512kb-----19,474 mb/s
1mb-----2,985 mb/s
 

fkloster

Diamond Member
Dec 16, 1999
4,171
0
0
system now running 1:1 @ 266fsb now that I removed 1 GB of my ram.... take note! I guess if you want to run fast ram on the i875.....don't run 2 GB of it...
 

NFS4

No Lifer
Oct 9, 1999
72,636
47
91
Originally posted by: fkloster
system now running 1:1 @ 266fsb now that I removed 1 GB of my ram.... take note! I guess if you want to run fast ram on the i875.....don't run 2 GB of it...

Check out my o/c benchmarks in my thread fkloster :D
 

Dug

Diamond Member
Jun 6, 2000
3,469
6
81
Originally posted by: fkloster
Ok guys... I'm exhausted but here is the benchmarks....anything look about right?

6,650 3dMarks build 330

50,453 AquaMark3

Sandra Standard v. 2002.6.8.97

CPU Arithmetic Benchmark:

Dhrystone ALU = 7,421 Mips
Whetstone FPU = 6,496 Mflops

CPU Multimedia Benchmark:

Integer = 14,921 it/s
Floating Point = 22,662 it/s

Filesystem Benchmark :) :

Drive Index = 67,790 mb/s

Memory Bandwidth Benchmark:

Ram Int Buffered = 6,145 mb/s
Ram Float Buffered = 6,173 mb/s

Cache & Memory Benchmark:

2kb-----34,204 mb/s
4kb-----34,978 mb/s
8kb-----28,355 mb/s
16kb-----24,797 mb/s
32kb-----25,366 mb/s
64kb-----25,381 mb/s
128kb-----25,808 mb/s
256kb-----23,542 mb/s
512kb-----19,474 mb/s
1mb-----2,985 mb/s


Can you please get up to date :)

Newest version of Sandra

Those numbers should be completely different.
 

bluemax

Diamond Member
Apr 28, 2000
7,182
0
0
I'm making a note to re-read this thread once I get my Intel Retail Edge package. I'm gonna' need new RAM to go on it, though I highly doubt I'll be able to afford the super-expensive stuff you guys are getting. :Q :Q :eek:
 

fkloster

Diamond Member
Dec 16, 1999
4,171
0
0
Dug...none of the benchmarks work on my rig. I DL'd them from your link...installed the 2004...i see the benchmarking modules...when I execute them... I get N/A for my score...
 

Megatomic

Lifer
Nov 9, 2000
20,127
6
81
When you open the benchmark module press the F5 key, the benchmark will run then. SiSoft did this intentionally.
 

fkloster

Diamond Member
Dec 16, 1999
4,171
0
0
Originally posted by: Megatomic
When you open the benchmark module press the F5 key, the benchmark will run then. SiSoft did this intentionally.


Is this something new because it did not do this in the old modules...
 

Megatomic

Lifer
Nov 9, 2000
20,127
6
81
Originally posted by: fkloster
Originally posted by: Megatomic
When you open the benchmark module press the F5 key, the benchmark will run then. SiSoft did this intentionally.


Is this something new because it did not do this in the old modules...
Yes, it is a new "feature". Probably a measure to ensure more accurate test results. I think that this is a method to prevent benchmark scores from being affected by inadvertant mouse movement and other performance obstacles.