Sandra's "first tests" are inflated

meowChow · 08-28-2003, 04:39 AM

After a fresh boot on my P4P800, I've noticed that my first run of the Sandra mem bandwidth benchmark (either buffered or unbuffered) is consistently about 2-5% higher than successive invocations.

Does anyone else experience this anomalous behavior?

Scoundrel · 08-28-2003, 04:46 AM

yep same here.

I think its normal, a number of other user s have reported this before.

meowChow · 08-28-2003, 04:47 AM

Exclusive to P4P800 or Asus boards, or just a Sandra thing?

Scoundrel · 08-28-2003, 04:54 AM

A sandra thing. I have P4T533-C, P4TE, IC7, P4S533-X, TUSL2-C and all fluctuate between 2-5%

chuckbam · 08-31-2003, 05:42 PM

As long as everyone runs the same test the same, whats the difference. If one run is a better score, probably eveyone uses that test. If thats the case, you need to run the first test to compare to other systems and reviews that use that mothed. I run tests after a fresh boot. I want to have a clean test with few TSRs and other Windows junk going on.

Miles · 08-31-2003, 06:36 PM

Hi meowChow,

Most review sites are publishing the Sandra results after a reboot and considering those to be accurate and subsequent results without a reboot to be deflated and inaccurate.

Smurfer · 08-31-2003, 07:09 PM

The older version worked just the opposite the more you ran the benchmark the higher the mem score.

meowChow · 09-01-2003, 03:51 AM

Quote:

Originally posted by Paragon
Hi meowChow,

Most review sites are publishing the Sandra results after a reboot and considering those to be accurate and subsequent results without a reboot to be deflated and inaccurate.

Interesting. My feeling about this is just the opposite. I believe reproducable result should be used. Since second run and subsequent runs produce similar results, those should be used. The first run result seems is probably erroneous, not vice-versa.

chuckbam · 09-01-2003, 04:28 AM

Quote:

Originally posted by meowChow
Interesting. My feeling about this is just the opposite. I believe reproducable result should be used. Since second run and subsequent runs produce similar results, those should be used. The first run result seems is probably erroneous, not vice-versa.

Well you have a different view than I do. I think a fresh boot is needed. Everyone has to test the same way for a test to be any good.

meowChow · 09-01-2003, 05:40 AM

Quote:

Originally posted by chuckbam
Well you have a different view than I do. I think a fresh boot is needed. Everyone has to test the same way for a test to be any good.

I agree, a fresh boot is needed, and everybody should do their tests the same way, but if everyone is using the first test score, then imho, everybody is doing it wrong.

chuckbam · 09-01-2003, 10:24 AM

Quote:

Originally posted by meowChow
I agree, a fresh boot is needed, and everybody should do their tests the same way, but if everyone is using the first test score, then imho, everybody is doing it wrong.

Does it make a difference in the unbuffered test?

StefanPI · 09-01-2003, 10:33 AM

Quote:

Originally posted by meowChow
After a fresh boot on my P4P800, I've noticed that my first run of the Sandra mem bandwidth benchmark (either buffered or unbuffered) is consistently about 2-5% higher than successive invocations.

Does anyone else experience this anomalous behavior?

I had the opposite problem on my p4s8x, 2.4ghz, ddr333. First 1-3 tests was low and about the 5th and onward was higher. Could be some bg service that interfered.

meowChow · 09-01-2003, 02:00 PM

Quote:

Originally posted by chuckbam
Does it make a difference in the unbuffered test?

Yes, the first-run affects both tests about the same amount, percentage-wise.

griffpa · 09-01-2003, 02:15 PM

Ive noticed the Sandra score fluctuate, but the first run isn't always the highest. I started running the test three times and taking the highest scores.

**specmike** · 09-01-2003, 02:21 PM

Quote:

Originally posted by meowChow
I agree, a fresh boot is needed, and everybody should do their tests the same way, but if everyone is using the first test score, then imho, everybody is doing it wrong.

How and why is there a "wrong" way to run a benchmark? That seems like a strange comment to me.

The point is to get the highest score you can by tuning and optimizing your computer. We all have access to the same version of Sandra and therefore the playing field is level.

Mike

08-28-2003, 04:39 AM	#1
meowChow Registered User Join Date: Jul 2003 Posts: 46	Sandra's "first tests" are inflated After a fresh boot on my P4P800, I've noticed that my first run of the Sandra mem bandwidth benchmark (either buffered or unbuffered) is consistently about 2-5% higher than successive invocations. Does anyone else experience this anomalous behavior?
(Offline)

08-28-2003, 04:46 AM	#2
Scoundrel Ferrari....hah!! Join Date: Jun 2001 Location: Copenhagen, Denmark Posts: 2,464	yep same here. I think its normal, a number of other user s have reported this before. __________________ Processor Intel Core2 Quad Q6600 G0@3.2Ghz Motherboard Gigabyte GA-X38T-DQ6 Memory 4GB DDR3 Corsair Twin3X4096-1600C7DHX G Graphics XFX Geforce 9800GX2 1Gig Audio X-FI Extreme Gamer Professional Harddrives 2x 150GB WD Raptor X in RAID 0 Optical Drives 2x LiteOn 20x SATA DVD/RW's Powersupply TT ToughPower 750W Case Coolermaster Stacker 830 Cooling Intel Box Cooler, & 3x Enermax Warp 120mm Fans Input Devices Logitech G5, G15, G25 Monitors 2x Samsung 226BW 22"
(Offline)

08-28-2003, 04:47 AM	#3
meowChow Registered User Join Date: Jul 2003 Posts: 46	Exclusive to P4P800 or Asus boards, or just a Sandra thing?
(Offline)

08-28-2003, 04:54 AM	#4
Scoundrel Ferrari....hah!! Join Date: Jun 2001 Location: Copenhagen, Denmark Posts: 2,464	A sandra thing. I have P4T533-C, P4TE, IC7, P4S533-X, TUSL2-C and all fluctuate between 2-5% __________________ Processor Intel Core2 Quad Q6600 G0@3.2Ghz Motherboard Gigabyte GA-X38T-DQ6 Memory 4GB DDR3 Corsair Twin3X4096-1600C7DHX G Graphics XFX Geforce 9800GX2 1Gig Audio X-FI Extreme Gamer Professional Harddrives 2x 150GB WD Raptor X in RAID 0 Optical Drives 2x LiteOn 20x SATA DVD/RW's Powersupply TT ToughPower 750W Case Coolermaster Stacker 830 Cooling Intel Box Cooler, & 3x Enermax Warp 120mm Fans Input Devices Logitech G5, G15, G25 Monitors 2x Samsung 226BW 22"
(Offline)

08-31-2003, 05:42 PM	#5
chuckbam Personal Advocate Join Date: Mar 2003 Location: Louisiana USA Posts: 2,284	As long as everyone runs the same test the same, whats the difference. If one run is a better score, probably eveyone uses that test. If thats the case, you need to run the first test to compare to other systems and reviews that use that mothed. I run tests after a fresh boot. I want to have a clean test with few TSRs and other Windows junk going on. __________________ i7 920 3.60 GHz, QPI. 6.47 GT/s, 1.22V ~ Noctua NH-U12P Cooler ~ Asus P6T Deluxe, Intel X58 Express/ ICH10R ~ 12GB Crucial D9JNM DDR3 1440 MHz, 8-8-8-20 1T, 1.58V ~ 3x MaXLine Pro 500 RAID 0 ~ Barracuda 7200.11 1TB ~ Two ATI Radeon RV670 GPU, 512MB DDR3 ~ Two ATI Theater 650 Pro ~ Silverstone OP650 PS ~ LIAN LI Mid Tower ~ 3 LCD Multiple Display w/ full custom Aero ... Vista Ultimate 64-bit ... System Runs 24/7 .... system_01
(Offline)

08-31-2003, 06:36 PM	#6
Miles Registered User Join Date: May 2001 Location: South Carolina Posts: 5,558	Hi meowChow, Most review sites are publishing the Sandra results after a reboot and considering those to be accurate and subsequent results without a reboot to be deflated and inaccurate. __________________ Regards, Miles ________________________________ Intel Core 2 Duo E6850 w/ Enzotech Ultra \| ASUS Blitz Formula SE \| 2 x 2Gb Mushkin XP2-6400 DDR-2 \| XFX 8800 GTX XXX\| 2 - Seagate 750GB Barracuda ES \| PCP&C 1KWSR \| Lian Li G70 \| Windows Vista 64-bit Ultimate \| Dell 2407 FPW Monitor
(Offline)

08-31-2003, 07:09 PM	#7
Smurfer Land of the Brave Join Date: Oct 2002 Location: W. Los Angeles, California Posts: 1,854	The older version worked just the opposite the more you ran the benchmark the higher the mem score.
(Offline)

09-01-2003, 02:15 PM	#14
griffpa Registered User Join Date: Jul 2003 Posts: 120	Ive noticed the Sandra score fluctuate, but the first run isn't always the highest. I started running the test three times and taking the highest scores.
(Offline)

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode