Cheesy memory slows down prime95?

nomadicus · 2003-02-26, 13:25

I have an Athlon XP 1700. I had 3 sticks of memory -- one of which was bad (discovered in the torture test -- thank goodness). The company that built the PC replaced the stick with a no-name brand (can't remember the name of the first two sticks but it was something I recognized).

When I installed the replacement stick my iteraction times went from .238/iteration to .280./interation. About 18% decrease in speed. So I took out the third stick and got back my .238/iteration time (on a 10M digit candidate). I'll handle the replacing the slow stick shortly; due to additional problems, I am never buying from this company again. What a difference decent memory makes. :)

Lesson learned is don't skimp on memory. Get high quality.

In the future I am going to build some P4 systems. I am pretty sure I am getting an Intel MB with a 533 FSB. Should the memory be Crucial? 333DDR RDRAM? Recommendations?

Thanks!
-=- john

QuintLeo · 2003-02-26, 16:08

I've had good luck pushing recent Kingston DDR past it's ratings - current PC2100 512 Meg stick (rated DDR266) is running 100% rock solid at 315ish.

I don't know if it's the CPU (2.0a at 2.36) limiting the FSB or the RAM - I think it's the CPU, though, as I had to back it off a hair from 2.38 to get Prime to run 100% reliable, and I think that was BEFORE I started working with the memory settings.

outlnder · 2003-02-26, 19:04

Try to always go a series higher. So if you need PC2700, get PC3200. This will allow a good overclick with keeping the "1T" command in the Bios. Since there is very little difference in price, the faster Ram will serve you better in the long run.

As for brand, Kingston is good. I don't buy Crucial(Micron) anymore due to numerous failures. Corsair XMS is supposed to be excellent, but don't buy the in between steps, ie PC3000.

ebx · 2003-02-26, 19:57

Buying a step up is good but has to wait for finalized standards.

I had good luck with Crucial and Corsiar(no overclock though). I burnt a memtest86 bootable CD and always run it overnight on any new memory. It reports the speed of L1/L2/memory as well.

Have to get quality memory, quality cpu, quality video, quality PS, quality case/fan, quality LAN, quality screen. I guess only thing left out is my mousepad. ;)

Tasuke · 2003-02-26, 21:07

http://www.ratpadz.com/

My 2 cents.

I do a lot of cad work and gaming. I have converted the other engineer as well. Do not get the logo one, no end of mess.

Or if you use a trackball(Outlnder) you do not need a mouse pad. FYI.

QuintLeo · 2003-02-26, 23:06

I never did understand why Xerox turned a perfectly good trackball upside down to create the original mouse for the Star.

Logitech Marbles are the way!

9-)

rectifire · 2003-02-27, 06:29

nomadicus:

Just thought I'd go into detail a little more on your problem and explain why. Many already know this and I apologize in advance for boring the majority of you who know this already.

Each memory stick has something on it called serial presence detect (SPD for short.) What the SPD does is tell the motherboard what settings the RAM is capable of running at. (example: Memory speed DDR266, 333, 400 Memory CAS level 2.5, 2 etc.) The motherboard then obeys these settings and configures itself to run the RAM at these specified values.

In case of multiple sticks of RAM, the motherboard will look at the SPD of each stick of RAM to see what each particular stick of RAM is capable of running at. If one stick of RAM has it's SPD programed with slow settings, the motherboard then must run ALL RAM at these slow settings. In other words, the RAM on your computer runs only as fast as the slowest stick.

This is what is happening in your case. When you insert the cheaper and slower no name brand stick of RAM into your computer, it forces all the RAM in your computer to run at a slower setting. This in turn causes your prime95 iteration times to increase, because memory performance is lower.

In short, quality name brand RAM rated to run at fast settings is a good thing. My guess is if you were to buy RAM of equal or better quality than your existing 2 sticks of RAM, there should be no difference in your iteration times.

adpowers · 2003-02-27, 06:55

I have had good luck with Kingston.

nomadicus · 2003-02-27, 18:17

Quote:

Originally Posted by rectifire

What the SPD does is tell the motherboard what settings the RAM is capable of running at. (example: Memory speed DDR266, 333, 400 Memory CAS level 2.5, 2 etc.)

That was a great piece of info for me. Thanks rectifire :)
Now I can be more intelligent when I talk to the place where I got it.

ebx · 2003-03-01, 00:15

The most important piece of information in SPD is how the memory is organized, in rows x columns x banks. That will decide how the sdram controller(in your north bridge) is programmed and thereafter, how the memory is accessed. This the the memory automatic detect by the bios. Many memory chips are capable to run at a few different speed and latency settings so those are really things CPU tells the chips to do. Bios will assume some safe default settings.

When it comes to actual use of memory, as long as the chip can understand the command(reads and writes), it does what it is told. If it is slow, the CPU has no choice other than wait, providing the wait is within the bus specification.

There are a few other numbers that cpu has to decide, such as delays between various commands. Other things include ECC and registered or not. The datasheet of the memory chips will tell the story.

2003-02-26, 13:25	#1
nomadicus Jan 2003 North Carolina 2·3·41 Posts	Cheesy memory slows down prime95? I have an Athlon XP 1700. I had 3 sticks of memory -- one of which was bad (discovered in the torture test -- thank goodness). The company that built the PC replaced the stick with a no-name brand (can't remember the name of the first two sticks but it was something I recognized). When I installed the replacement stick my iteraction times went from .238/iteration to .280./interation. About 18% decrease in speed. So I took out the third stick and got back my .238/iteration time (on a 10M digit candidate). I'll handle the replacing the slow stick shortly; due to additional problems, I am never buying from this company again. What a difference decent memory makes. :) Lesson learned is don't skimp on memory. Get high quality. In the future I am going to build some P4 systems. I am pretty sure I am getting an Intel MB with a 533 FSB. Should the memory be Crucial? 333DDR RDRAM? Recommendations? Thanks! -=- john

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Prime95 P-1 memory leak?	TheMawn	Software	5	2013-12-12 01:22
i7/prime95 memory question	stars10250	Hardware	115	2012-10-24 15:24
Prime95 v25.7 memory usage	opyrt	Software	1	2008-11-09 08:43
prime95 slows computer after I stop	Raterus	Software	7	2006-08-14 12:42
Prime95 Memory Usage	jimmyhua	Software	7	2005-07-10 07:37

2003-02-26, 16:08	#2
QuintLeo Oct 2002 Lost in the hills of Iowa 2⁶×7 Posts	I've had good luck pushing recent Kingston DDR past it's ratings - current PC2100 512 Meg stick (rated DDR266) is running 100% rock solid at 315ish. I don't know if it's the CPU (2.0a at 2.36) limiting the FSB or the RAM - I think it's the CPU, though, as I had to back it off a hair from 2.38 to get Prime to run 100% reliable, and I think that was BEFORE I started working with the memory settings.

2003-02-26, 19:04	#3
outlnder Aug 2002 100111110₂ Posts	Try to always go a series higher. So if you need PC2700, get PC3200. This will allow a good overclick with keeping the "1T" command in the Bios. Since there is very little difference in price, the faster Ram will serve you better in the long run. As for brand, Kingston is good. I don't buy Crucial(Micron) anymore due to numerous failures. Corsair XMS is supposed to be excellent, but don't buy the in between steps, ie PC3000.

2003-02-26, 19:57	#4
ebx Aug 2002 145₈ Posts	Buying a step up is good but has to wait for finalized standards. I had good luck with Crucial and Corsiar(no overclock though). I burnt a memtest86 bootable CD and always run it overnight on any new memory. It reports the speed of L1/L2/memory as well. Have to get quality memory, quality cpu, quality video, quality PS, quality case/fan, quality LAN, quality screen. I guess only thing left out is my mousepad. ;)

2003-02-26, 21:07	#5
Tasuke Aug 2002 72₁₀ Posts	http://www.ratpadz.com/ My 2 cents. I do a lot of cad work and gaming. I have converted the other engineer as well. Do not get the logo one, no end of mess. Or if you use a trackball(Outlnder) you do not need a mouse pad. FYI.

2003-02-26, 23:06	#6
QuintLeo Oct 2002 Lost in the hills of Iowa 2⁶×7 Posts	I never did understand why Xerox turned a perfectly good trackball upside down to create the original mouse for the Star. Logitech Marbles are the way! 9-)

2003-02-27, 06:29	#7
rectifire Feb 2003 10₂ Posts	nomadicus: Just thought I'd go into detail a little more on your problem and explain why. Many already know this and I apologize in advance for boring the majority of you who know this already. Each memory stick has something on it called serial presence detect (SPD for short.) What the SPD does is tell the motherboard what settings the RAM is capable of running at. (example: Memory speed DDR266, 333, 400 Memory CAS level 2.5, 2 etc.) The motherboard then obeys these settings and configures itself to run the RAM at these specified values. In case of multiple sticks of RAM, the motherboard will look at the SPD of each stick of RAM to see what each particular stick of RAM is capable of running at. If one stick of RAM has it's SPD programed with slow settings, the motherboard then must run ALL RAM at these slow settings. In other words, the RAM on your computer runs only as fast as the slowest stick. This is what is happening in your case. When you insert the cheaper and slower no name brand stick of RAM into your computer, it forces all the RAM in your computer to run at a slower setting. This in turn causes your prime95 iteration times to increase, because memory performance is lower. In short, quality name brand RAM rated to run at fast settings is a good thing. My guess is if you were to buy RAM of equal or better quality than your existing 2 sticks of RAM, there should be no difference in your iteration times.

2003-02-27, 06:55	#8
adpowers Sep 2002 1010000₂ Posts	I have had good luck with Kingston.

2003-03-01, 00:15	#10
ebx Aug 2002 101 Posts	The most important piece of information in SPD is how the memory is organized, in rows x columns x banks. That will decide how the sdram controller(in your north bridge) is programmed and thereafter, how the memory is accessed. This the the memory automatic detect by the bios. Many memory chips are capable to run at a few different speed and latency settings so those are really things CPU tells the chips to do. Bios will assume some safe default settings. When it comes to actual use of memory, as long as the chip can understand the command(reads and writes), it does what it is told. If it is slow, the CPU has no choice other than wait, providing the wait is within the bus specification. There are a few other numbers that cpu has to decide, such as delays between various commands. Other things include ECC and registered or not. The datasheet of the memory chips will tell the story.