Sandra's "first tests" are inflated - Page 2

chuckbam · 09-01-2003, 10:05 PM

I just did Sandra 2x and had 100% same results. I think this thread is becoming a waste.

I going to post my stuff..

meowChow · 09-01-2003, 10:24 PM

Quote:

Originally posted by chuckbam
I just did Sandra 2x and had 100% same results. I think this thread is becoming a waste.

The fact that Sandra is "consistent" on your system and inconsistent on others is even more cause for alarm. I'm not talking about a middling difference in first test scores, I've seen scores fluctuate 200+ pts.

It's just more reason to eschew Sandra benches as a gauge of system / memory performance.

chuckbam · 09-01-2003, 10:32 PM

Quote:

Originally posted by meowChow
The fact that Sandra is "consistent" on your system and inconsistent on others is even more cause for alarm. I'm not talking about a middling difference in first test scores, I've seen scores fluctuate 200+ pts.

It's just more reason to eschew Sandra benches as a gauge of system / memory performance.

I think that many overclockers think that MHz is everything. Often they run very unstable poor running systems. Maybe that has something to do with it.

I have been on a few systems that had .30MHz+ more then my system that seemed much slower when I was on them.

**specmike** · 09-01-2003, 11:53 PM

LOL....good lookin system there Chuck. You look like a RAM vendor.

oldfart · 09-01-2003, 11:56 PM

It's very easy to fix this Sandra problem. I did it long ago.

Start/settings/ control panel/add-remove programs......

chuckbam · 09-01-2003, 11:56 PM

Quote:

Originally posted by specmike
LOL....good lookin system there Chuck. You look like a RAM vendor.

Thanks and nope, just a system lover

chuckbam · 09-02-2003, 01:26 AM

Quote:

Originally posted by specmike
LOL....good lookin system there Chuck. You look like a RAM vendor.

specmike:
Have you ever tested you memory settings at Cas 2 3-2-5 ?

**k0NG0** · 09-02-2003, 03:30 AM

Well, my initial tests indicated that Sandra by no means is an absolute benchmark. I did 15-20 runs with different memory fragmentations but with the same configuration (processes started, timings, frequency etc.) and got a 3% variance in the scores (i.e. all scores were inside 3% of the mean score). The numbers were starting to converge, but this is enough for me to draw a conclusion that Sandra is by no means consistent in what score it spits out at the end.

My point with this test was actually to show that benchmarkers using a 1% or 2% increase as a sign of superior performance need to rethink what they write. There's too much noise in the measurements to accept any score within a certain margin of error - my pet project is to do a significant number of runs of different benchmarks to see how well they actually do in terms of accuracy between runs.

-kONGO

**specmike** · 09-02-2003, 12:13 PM

Quote:

Originally posted by chuckbam
specmike:
Have you ever tested you memory settings at Cas 2 3-2-5 ?

Not with this RAM Chuck. I have gone as low as 2.5,3,4,5,8 but it was not stable there. It will run 2.5,3,4,7,8 at 265/530 though.

Mike.

chuckbam · 09-02-2003, 09:03 PM

Quote:

Originally posted by specmike
Not with this RAM Chuck. I have gone as low as 2.5,3,4,5,8 but it was not stable there. It will run 2.5,3,4,7,8 at 265/530 though.

Mike.

Just thought I would ask.

johnrr6 · 09-02-2003, 09:50 PM

Shouldn't be taken as an end all beat all----just another indicator-----although I think the UNbuffered is a much better indicator than buffered.

SuperPi I think probably gives a better indicator of "what's faster"

Even the old standbys 3dMark 2001 and 2003 give an OK indication----but again----not the mecca of truth especially with all the "optomized driver" shennanigans going on.

Another good test is complicated video rendering. See how fast your system can render the same video clip.

I HAVE noticed that most good testers use a vast array of benches designed for different types of systems:
Business
Gamer
Normal, everyday

A whole bunch of tools in the toolbox.

Me----I'm easy-----whichever one shows my system faster-----hey, thats the one that I'll swear by.

John

johnrr6 · 09-02-2003, 09:55 PM

Off topic-----Did you have to overclock the heck out of the 5900----or just a wee bit??

We have similar CPUs-----I'd have to push my 9800pro pretty hard to get 6500.

John

chuckbam · 09-02-2003, 10:09 PM

Quote:

Originally posted by johnrr6
Off topic-----Did you have to overclock the heck out of the 5900----or just a wee bit??

We have similar CPUs-----I'd have to push my 9800pro pretty hard to get 6500.

John

Dear johnrr6 :

It is a eVGA card and they shipped overclocked at 500/900MHZ.
It is a referance model made by Flextronics for NVIDA. They shipped it to eVGA for boxing etc. It uses Hynix BGA 2.2ns memory. The memory is rated to 900MHZ. Thus, all basic referance models should run at this speed. Most Ultras are clocked at 450/850MHz.

The card will not run well much over the 500/900MHz

johnrr6 · 09-02-2003, 10:58 PM

Did you get all the little EVGA bugs worked out??

I remember a while back----there was some problems with their initial shipments.

But I REALLY like EVGA-----great company from my experience.

John

chuckbam · 09-02-2003, 11:20 PM

Quote:

Originally posted by johnrr6
Did you get all the little EVGA bugs worked out??

I remember a while back----there was some problems with their initial shipments.

But I REALLY like EVGA-----great company from my experience.

John

I have $30.00 worth of bucks, but their cards are so expensive, not much I can do with them.

I did not get excided over the whole deal. And shortly, they provided the referance bios and all was fine. I now have their new performance bios.

I think lots of people rushed out to get the card (before it was even officialy relased, and started lots of drama.

I was one of the first to get it also. I had fun for a week or two fooling with it. I did not looose the faith.

It helped to know I could retun it.

09-01-2003, 10:05 PM	#16
chuckbam Personal Advocate Join Date: Mar 2003 Location: Louisiana USA Posts: 2,285	I just did Sandra 2x and had 100% same results. I think this thread is becoming a waste. I going to post my stuff.. __________________ i7 920 3.60 GHz, QPI. 6.47 GT/s, 1.25V ~ 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 Last edited by chuckbam : 09-01-2003 at 10:17 PM.
(Offline)

09-01-2003, 11:53 PM	#19
specmike Yes, Yellowbeard Join Date: Oct 2002 Location: Metro Atlanta, GA. Posts: 2,627	LOL....good lookin system there Chuck. You look like a RAM vendor. __________________ CORSAIR TECH SUPPORT
(Offline)

09-01-2003, 11:56 PM	#20
oldfart Future Half-Life 2 Junkie Join Date: Aug 2001 Posts: 271	It's very easy to fix this Sandra problem. I did it long ago. Start/settings/ control panel/add-remove programs...... __________________ Abit IC7 2.6C @ 257 FSB 3.34 GHz 1.6 Vcore Swiftech MCX4000 \| Panaflo M1A 2 x 512 XMS3200C2 @ DDR412 2-2-2-6 WD 1200JB Lite-On DVD Pioneer A05 DVD-R/W Radeon 9700P \| Arctic Cooling Silencer TBSC Panaflo L1A case fan Antec True330 Win XP
(Offline)

09-02-2003, 03:30 AM	#23
k0NG0 Remembering TQ Join Date: Mar 2001 Location: Sweden Posts: 13,638	Well, my initial tests indicated that Sandra by no means is an absolute benchmark. I did 15-20 runs with different memory fragmentations but with the same configuration (processes started, timings, frequency etc.) and got a 3% variance in the scores (i.e. all scores were inside 3% of the mean score). The numbers were starting to converge, but this is enough for me to draw a conclusion that Sandra is by no means consistent in what score it spits out at the end. My point with this test was actually to show that benchmarkers using a 1% or 2% increase as a sign of superior performance need to rethink what they write. There's too much noise in the measurements to accept any score within a certain margin of error - my pet project is to do a significant number of runs of different benchmarks to see how well they actually do in terms of accuracy between runs. -kONGO __________________ Use Firefox - "the one that blocks all the schmutz" Feeling multicore elation? Remember this correlation: Amdahl's Law.
(Offline)

09-02-2003, 09:50 PM	#26
johnrr6 Lucky Amateur Join Date: Apr 2003 Location: Virginia Posts: 2,426	Sandra's just another tool in the toolbox Shouldn't be taken as an end all beat all----just another indicator-----although I think the UNbuffered is a much better indicator than buffered. SuperPi I think probably gives a better indicator of "what's faster" Even the old standbys 3dMark 2001 and 2003 give an OK indication----but again----not the mecca of truth especially with all the "optomized driver" shennanigans going on. Another good test is complicated video rendering. See how fast your system can render the same video clip. I HAVE noticed that most good testers use a vast array of benches designed for different types of systems: Business Gamer Normal, everyday A whole bunch of tools in the toolbox. Me----I'm easy-----whichever one shows my system faster-----hey, thats the one that I'll swear by. John __________________ 1st Rig:------------------------------------------------------2nd Rig: i7 920 @ 3.6 Ghz--------------------------------------------9850 BE at 2.6 Ghz Vista Ultimate Edition 64 Bit SP1---------------------------Vista Ultimate Edition 64 Bit SP1 Asus P6T Deluxe, BIOS 1101------------------------------AsRock 780G BIOS 1.50 Vigor Monsoon III LT---------------------------------------Thermaltake Blue Orb II 6 x 2 Gig Mushkin PC12800--------------------------------Mushkin 4 x 2 Gig PC8500 ATI HD 4870 1 Gig------------------------------------------ATI HD 4870 2 x 500 Gig Seagate SATA II HDs in RAID 0--------------2 x 500Gig Western Digital SATA II HDs in RAID 0 Samsung SATA TruDirect DVD Burner w/Lightscribe-------Samsung SATA Tru Direct DVD Burner w/Lightscribe Seasonic 600W PSU-----------------------------------------Silverstone 500 Watt ST50EF PSU Creative X-Fi Titanium sound-------------------------------Azalia Onboard Sound Lian Li AO5B Black Case------------------------------------Silverstone SG01 SFF Black Case Hauppage 1800 TV Card------------------------------------Hauppage 1600 TV Card
(Offline)

09-02-2003, 09:55 PM	#27
johnrr6 Lucky Amateur Join Date: Apr 2003 Location: Virginia Posts: 2,426	Chuck---pretty nice 2003 score Off topic-----Did you have to overclock the heck out of the 5900----or just a wee bit?? We have similar CPUs-----I'd have to push my 9800pro pretty hard to get 6500. John __________________ 1st Rig:------------------------------------------------------2nd Rig: i7 920 @ 3.6 Ghz--------------------------------------------9850 BE at 2.6 Ghz Vista Ultimate Edition 64 Bit SP1---------------------------Vista Ultimate Edition 64 Bit SP1 Asus P6T Deluxe, BIOS 1101------------------------------AsRock 780G BIOS 1.50 Vigor Monsoon III LT---------------------------------------Thermaltake Blue Orb II 6 x 2 Gig Mushkin PC12800--------------------------------Mushkin 4 x 2 Gig PC8500 ATI HD 4870 1 Gig------------------------------------------ATI HD 4870 2 x 500 Gig Seagate SATA II HDs in RAID 0--------------2 x 500Gig Western Digital SATA II HDs in RAID 0 Samsung SATA TruDirect DVD Burner w/Lightscribe-------Samsung SATA Tru Direct DVD Burner w/Lightscribe Seasonic 600W PSU-----------------------------------------Silverstone 500 Watt ST50EF PSU Creative X-Fi Titanium sound-------------------------------Azalia Onboard Sound Lian Li AO5B Black Case------------------------------------Silverstone SG01 SFF Black Case Hauppage 1800 TV Card------------------------------------Hauppage 1600 TV Card
(Offline)

09-02-2003, 10:58 PM	#29
johnrr6 Lucky Amateur Join Date: Apr 2003 Location: Virginia Posts: 2,426	Nice Did you get all the little EVGA bugs worked out?? I remember a while back----there was some problems with their initial shipments. But I REALLY like EVGA-----great company from my experience. John __________________ 1st Rig:------------------------------------------------------2nd Rig: i7 920 @ 3.6 Ghz--------------------------------------------9850 BE at 2.6 Ghz Vista Ultimate Edition 64 Bit SP1---------------------------Vista Ultimate Edition 64 Bit SP1 Asus P6T Deluxe, BIOS 1101------------------------------AsRock 780G BIOS 1.50 Vigor Monsoon III LT---------------------------------------Thermaltake Blue Orb II 6 x 2 Gig Mushkin PC12800--------------------------------Mushkin 4 x 2 Gig PC8500 ATI HD 4870 1 Gig------------------------------------------ATI HD 4870 2 x 500 Gig Seagate SATA II HDs in RAID 0--------------2 x 500Gig Western Digital SATA II HDs in RAID 0 Samsung SATA TruDirect DVD Burner w/Lightscribe-------Samsung SATA Tru Direct DVD Burner w/Lightscribe Seasonic 600W PSU-----------------------------------------Silverstone 500 Watt ST50EF PSU Creative X-Fi Titanium sound-------------------------------Azalia Onboard Sound Lian Li AO5B Black Case------------------------------------Silverstone SG01 SFF Black Case Hauppage 1800 TV Card------------------------------------Hauppage 1600 TV Card
(Offline)

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