RAM question for Giga-byte GA-5AX rev 2. (Update 7-9-06)

Discussion relating to Socket 7 hardware.
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

Jim, thanks for asking KGB! ;)

I was just about to ask the same, albeit I would like him to use 3x128mb sticks if available.

as I explained in a post at the beginning of this thread I suddenly realized that my write-scores in everest had fallen to a mere 76mb. at that time I was using 2 to 3 256mb sticks. I was really upset because I know that once I reached 120 to 130 on that machine (without any tweaks). unfortunately I didn't log these results. I am almost certain but can't say for sure if that was before I upgraded the system with those 256mb sticks.

so I went back to only 1 256mb stick and voila, I was back at around 120. as soon as I put in an additional other stick - no matter what size - I am back to 76. so I went the good old route put in 3x128mb - the write score is back to 120. maybe that this is a sign that there is a general problem with 256-sticks and above .... :( (these tests were done with a k6-2+550)

anyway, I am still wondering how you managed to get these great read-results without software tweaking.

you wrote that at the setting 5.5x100, 512mb ram, cache=off, bypass, timing fastest, cas=2, 2/2 <- what did you mean with that!?? is this some setting in the bios?
you got everest read of 374! that is 18 more than I get with a k6-2-550 (no plus-sign).

my read score of 356 is with memory timings of 2-2-4-5-8. this info I got from everest. what was (is) yours?

I have to manually (wpcredit) set the timings to get 2-2-2-4-7 and with 2 more wpcredit-tweaks get finally a read score of 367, still 7 below your 374 :(

so let me summarize my questions:
what does everest report as your memory-timings?
what do you mean by "2/2"
do you get better write-scores when using a smaller ram-stick (256 or lower)?

many thanks if you could dig again into this!
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

thanks Jim for the hint on "rls48j" .... but I doubt the 1.06 can carry a "+"-cpu even at acceptable speed. I already tried some time ago and I experienced what so many owners of a 1.06 board suffered - very poor performance.

I did not dig further into this because the information I had gathered so far about this issue was/is that there is no remedy :(

so if you learned that "rls48j" succeeded in running that combo I will give it another try (sooner or later).
User avatar
KGB
K6'er
Posts: 98
Joined: Mon Apr 07, 2003 7:55 pm

Post by KGB »

You guys know just how to make an itch turn into a rash :wink:

With the Asus being RMA'd, without any clear idea of when it will return :oops: I will have to grab my screw driver and go back into the abyss..

Just when I thought I was out, you guys push me back in :shakes fists into air: DAMN YOU MANIACS DAAAAMN YOU :P

Jim, Don as you know, I need a platform to test this on. The really good Sandra memory numbers and arithmatic numbers came in Windows XP sp1. The high Everest read/write test results came in Windows 2000 Pro.

Don concerning your question, in the BIOS, where you set the CHIPSET FEATURES: The THIRD one down the list AT BUS Clock it has several options: 2/2-2/3-2/4-2/5 7.159 MHZ. Now, I don't know what this is, but I set it to 2/2. Hope that answer that, ill probably google myself, but let me know if you have any information on it.
-K6-2 550 Gigagyte GA-5AX (5.2) 1x512MB GeForce 2 Ti500 64MB
-Pentium III 933 Asus P3C-L 2x128MB GeForce 4 Ti4200 128MB DDR
-Pentium III 750 Asus P3B-F ATi Rage128 GL 32MB
-Celeron 1.2 Asus TUSI-M 2x128MB
-Pentium IV 1.5 DELL 2350 1x128MB
-Athlon 900 Asus K7M 1x256MB 2x128MB GeForce 3 64MB
-AthlonXP-M 2400+ Asus K7V880 2x512MB HIS Radeon HD3850 512MB DDR
User avatar
KachiWachi
K6'er Elite
Posts: 507
Joined: Wed Sep 21, 2005 10:53 am
Location: Pennsylvania, USA

Post by KachiWachi »

AT BUS Clock sets the speed of your AT Bus (ISA Bus). It should be set to 8.33 MHz...or as close as you can get to it.

The numbers you see - 2/2, 2/3, 2/4, 2/5 - are fractional dividers, while the 7.159 MHz is a fixed number.
Last edited by KachiWachi on Fri Mar 03, 2006 6:43 am, edited 1 time in total.
Moderator - Wim's BIOS

PC #1 - DFI 586IPVG, K6-2/+ 450 (Cyrix MII 433), 128 MB EDO. BIOS patched by Jan Steunebrink.
PC #2 - Amptron PM-7900 (M520), i200 non-MMX, 128 MB EDO
PC #3 - HP8766C, PIII-667, 768 MB SDRAM
PC #4 - ASUS P3V4X, PIII-733, 256 MB SDRAM
PC #5 - Gateway 700X, P4-2.0 GHz, 768 MB PC800 RDRAM
PC #6 - COMPAQ Evo N1020v laptop, P4-2.4 GHz, 1 GB PC2700 DDR
PC #7 - Dell Dimension 4600i, P4-2.8 GHz, 512 MB PC2700 DDR
PC #8 - Acer EeePC netbook, Atom N270 @ 1.60 GHz, 1 GB RAM
PC #9 - ??? ;)
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

@ Peter. If you read very old posts here, you will find that "rls48j" DID MANAGE to get a K6-+ processor working properly on one of those boards. More recently "SaintlyCobra" was working w/ one of those boards trying to tweak for max performance, and Cake arranged for rls48j to help him. Unfortunately no sooner had everybody connected than SaintlyCobra announced that he had blown up the board by trying for 5x120. Suggest you contact rls48j by PM or ask Cake to get in touch w/ him.

Note: rls48j Did NOT flash his bios to the 1011, (probably the source of the slowness problem), but instead experimented for months w/ WPCredit to get it working properly.

@ Peter & KGB : I think that 2x256 is more appropriate than 3x128, because in the latter case you are changing the AMOUNT of RAM installed, not just the configuration. What I wanted to know is "Does the concentration of memory in one slot make a difference, when all else is equal?"
That is not to say it is inappropriate to test 3x128, that would just be a different test w/ more variables.

@ Peter. When using 2x256, or 3x256, did you try it w/ cache enabled? - and try running the Sandra test several times? - then running the Everest test after?

@ KGB. Thanks for going back into battle so soon after the last war ended, You are fighting for a good cause!!
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
User avatar
KachiWachi
K6'er Elite
Posts: 507
Joined: Wed Sep 21, 2005 10:53 am
Location: Pennsylvania, USA

Post by KachiWachi »

@RAM - Does this board Interleave? If the RAM chip density changes, this *might* change the Interleave, which *may* change memory test scores.
Moderator - Wim's BIOS

PC #1 - DFI 586IPVG, K6-2/+ 450 (Cyrix MII 433), 128 MB EDO. BIOS patched by Jan Steunebrink.
PC #2 - Amptron PM-7900 (M520), i200 non-MMX, 128 MB EDO
PC #3 - HP8766C, PIII-667, 768 MB SDRAM
PC #4 - ASUS P3V4X, PIII-733, 256 MB SDRAM
PC #5 - Gateway 700X, P4-2.0 GHz, 768 MB PC800 RDRAM
PC #6 - COMPAQ Evo N1020v laptop, P4-2.4 GHz, 1 GB PC2700 DDR
PC #7 - Dell Dimension 4600i, P4-2.8 GHz, 512 MB PC2700 DDR
PC #8 - Acer EeePC netbook, Atom N270 @ 1.60 GHz, 1 GB RAM
PC #9 - ??? ;)
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

@KachiWachi

I have an old everest version (1.52.511). it reports under computer/dmi/memory controller that the supported/current memory interleave is 1. but I have also learned that the dmi-information is something that can be altered arbitrarily by the mainboard manufacturer .... :( so it might not be accurate.
thanx for the info on the at-bus speed thing!


@Jim
thanks for all the info re p5a 1.06 :) I will undoubtely try everything in getting the most out of it.

re sandra: I experienced this cut in everest write performance with 2x256 only on the ga5ax so far. I didn't test this configuration with my p5a yet, but I will check it there out too.
re sandra#2: I am using sandra 2001-750 and this can also be the reason that it does not show the phenomenon of rising scores by multiple runs. but what I found out with the 2001version is that when I choose the benches in order cpu/mem/multimedia and I rerun the mem-test before the multimedia test I don't get improved results (+1/2 max). when I then run the multimedia test and immediately after it run the mem-test again the scores are up 10 to 12 for the int/alu score and about 1 to 2 for the float/alu score. another rerun thereafter does not improve results anymore.

but in general I would say that gathering data (bench results) this way is not very meaningful because it is in no way representative of actual performance. when does one run an application before/after an other application and then multiple times to max it out. I don't use any application that way. also such a behaviour of a bench tool reveals weakness of the tool in measuring what it is supposed to examine. how can you reliably compare results? what do they mean? is my sytem 10 or even 20% faster just because of benching it that way? sad thing .....


@ KGB:
with regards to KachiWachi's explanation you should set the AT-bus clock to clk2/4 (=33/4). Ithat was my configuration from the beginning. of course I could not resist to try 2/2 anyway but experienced no performance gain (would have been a miracle, but chipsets' ways are sometimes unpredictable. I would smear butter on the on/off knob if it would boost performance ;)
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

@ Peter : The point I am making is that where large amounts of RAM are uncached @ 3rd level, performance will improve due to increased cache hits as the application runs. The longer the application runs, the greater the percentage of cache hits, up to a point. This results in the application speeding up as it is running. That is as equally applicable to "Real World" applications as it is to benchmarks; and therefore IS representative of actual performance.

Think of it as being similar to Cosbi OS Mark's "Official Run", wherein each test is run 3 times; and the best result taken for the "Official Score"; except that my results are much more "Repeatable".

Be it noted, that if all of the same amount of RAM were cached @ 3rd level, the results would differ. Instead of rising, they would be flat; but at a higher level.
Last edited by Jim on Sat Mar 04, 2006 1:51 pm, edited 1 time in total.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
User avatar
KachiWachi
K6'er Elite
Posts: 507
Joined: Wed Sep 21, 2005 10:53 am
Location: Pennsylvania, USA

Post by KachiWachi »

That's an old Everest...

As to comparing your memory test results, make sure you all are using the same versions of the test software, as there *can* be differences in your results because of that.

Interleave can be found on the Motherboard/Chipset page in later versions of Everest. However, note that Everest only reads the last memory row to determine this (IIRC), so you need to check the chipset registers to see what each memory module is actually set for. As to the module capabilities, this is shown on the Motherboard/SPD page (under Module Size...rank/bank). The bank reading would be the Interleave capability of the module itself.
Moderator - Wim's BIOS

PC #1 - DFI 586IPVG, K6-2/+ 450 (Cyrix MII 433), 128 MB EDO. BIOS patched by Jan Steunebrink.
PC #2 - Amptron PM-7900 (M520), i200 non-MMX, 128 MB EDO
PC #3 - HP8766C, PIII-667, 768 MB SDRAM
PC #4 - ASUS P3V4X, PIII-733, 256 MB SDRAM
PC #5 - Gateway 700X, P4-2.0 GHz, 768 MB PC800 RDRAM
PC #6 - COMPAQ Evo N1020v laptop, P4-2.4 GHz, 1 GB PC2700 DDR
PC #7 - Dell Dimension 4600i, P4-2.8 GHz, 512 MB PC2700 DDR
PC #8 - Acer EeePC netbook, Atom N270 @ 1.60 GHz, 1 GB RAM
PC #9 - ??? ;)
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

@ KachiWachi ; You are right. The problem is that most people settle on one version of Everest and Sandra, so as to insure that their own results are comparable. e.g. I am currently using the most recent version of Everest that you posted a link to; but with Sandra, I use 2004 because that is what ALL my Sandra tests have been run with; and I want them to be comparable w/ each other. Taz uses 2003 probably for the same reason; and Peter uses 2001, most likely again for the same reason.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

MSI MS5169

Post by Stedman5040 »

Having read this thread through and through I thought I might give you some offerings from an as yet mentioned board the humble MSI MS5169. This board has been set up as follows

MS5169 Ver4.0 (ALI AladdinV)
K6-III+/450ACZ @ 550 (5.5x100)
3x256Mb high density PC133Ram (records 768Mb)
Intel i740 AGP
Creative SB128
Fujitsu 6.0Gig HDD
24x cdrom
Win ME

The reason I chose to run these benchmarks on this board was because it would support the high density SDRAM so I could really beef up the memory as per comments by JIM. I run ctu at start up to enable write allocate etc and without any further tweaks from wpcredit get the following from Everest and sandra2004 (Same as Jim).

Everest read/write/latency
333/156/200
Sandra memory bandwidth
190/193

Aladdin 5 chipsets do seem to give reasonable scores from the off.

By wpcredit tweaking the dram to give timings of 2-2-2-5 with everything else as default

everest scores give 334/163/200
Sandra gives 196/198

By wpcredit tweaking mixed dram command interval to 3T/5T we get

Everest

333/176/199

Sandra2004

198/196 rising to 208/206 after repeat tests

A further wpcredit tweak enabling fast dram read gives

Everest 351/183/191
Sandra 212/210

Considering that the fsb speed has been kept to 100mhz the scores do look quite respectable. Unfortunately the board does not have any settings for fsb's higher than 100. It would be nice to try this at higher fsb's. The clock chip on this board is an ic works 9248bf-131. Does anyone know if this chip can get clock speeds higher than 100fsb? I tried softfsb and it was no go above 100

As a comparison I did some benchmarks on an AX59PRO motherboard (VIA chipset) with again a K6-III+450ACZ @ 550 (5.5x100). Using all the seemingly endless tweaks in Jim's very useful thread on the DFI KBV3+ board I got the following results. Please bear in mind that the memory capacity was only 384Mb.

Everest

266/142/245

Sandra

180/178

Again this was using the same cpu at the same speed on this AX59 PRO setup. Results without any tweaks and just the ctu enabling features I got as follows

Everest

236/129/262

Thanks again to Jim as the tweaks on wpcredit are not easy to find.

I would try the same experiment on the P5A board that I have, but it does not take the high density ram so at the moment I would be stuck with 384MB.

Any suggestions welcome for further enhancements.

stedman
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Poor Memory Writes

Post by Stedman5040 »

@Donpedro

Whilst messing about with the memory on my MSI board my memory write scores in Everest plummeted to 78. This was really strange as the memory read scores and the memory latency scores were OK. I had installed two extra 256 dimms. Having a quick look round I found that the Write allocate memory size hadn't changed from the original 256 to 768. As soon as I changed the write allocate to 768 and rebooted the memory write scores were back up to where they should have been. Maybe this is the source of your memory problems with the 256mb sticks. :wink:

Stedman
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

RE: MSI MS5169

Post by Jim »

T H A N K - Y O U -- S T E D M A N !!!!
Whole lot to discuss here.
1) Go to http://www.cpu-cool.de/index.html and you will wind up at the creator of CPU Cool's site. Be prepared for a deluge of Adware. CPU Cool is a software program that will allow you to OC the FSB on that MSI. My DFIs don't have any provision to go over 100Mhz either, but I have done it w/ CPU Cool.

Will edit this to add more as soon as I get rid of some of the adware.
K, done that.

2) Is it possible that you missed the WPCredit tweak regards setting the "Dram Start Cycle" to "with cache @ 66 Mhz" as opposed to "after cache @ 100 Mhz" ? The "with cache @ 66 Mhz" setting gives one of the biggest gains available.

3) WinXP seems to generally give better results than Win98SE. Don't know about WinME because I have never had or used it.

4) Did you try re-running Everest immediately after getting your peak Sandra "Memory Bandwidth" scores? I get a big jump in my Everest scores doing that. I think it is because Everest functions in such a manner that it does not tend to generate a rising level of 3rd level cache hits; but Sandra does; and re-running Everest after the Sandra "Memory Bandwidth" test has peaked, somehow enables Everest to take advantage of an increased level of cache hits created by Sandra.
Last edited by Jim on Sat Mar 04, 2006 5:01 pm, edited 2 times in total.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
User avatar
KGB
K6'er
Posts: 98
Joined: Mon Apr 07, 2003 7:55 pm

Post by KGB »

Still tweaking with the 1 512mb memory modules guys:
Windows 2000 Stock is as usual as before

Windows 2000 Tweaked (K6speed, Wpcredit)
Everest:
395 read 8)
142 Write :shock:
215 Latency :o

Sandra is still testing :(

THANK YOU VERY MUCH JIM :D I appreciate your mini article on using cpcredit. I'll update this once I fine tune the system and make registry files.
-K6-2 550 Gigagyte GA-5AX (5.2) 1x512MB GeForce 2 Ti500 64MB
-Pentium III 933 Asus P3C-L 2x128MB GeForce 4 Ti4200 128MB DDR
-Pentium III 750 Asus P3B-F ATi Rage128 GL 32MB
-Celeron 1.2 Asus TUSI-M 2x128MB
-Pentium IV 1.5 DELL 2350 1x128MB
-Athlon 900 Asus K7M 1x256MB 2x128MB GeForce 3 64MB
-AthlonXP-M 2400+ Asus K7V880 2x512MB HIS Radeon HD3850 512MB DDR
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

Quite welcome! Glad to see the KGB has returned from the war victorious; having successfully driven back the enemy on all fronts. Especially "Memory Write"!!

Also am looking forward to more test results. i.e. Run the Sandra "Memory Bandwidth" test several times in succession to see if your results rise or stay flat. I.e. Do rising results still occur w/ mobo cache disabled.

Also would like you to check in WPCredit if your chipset's internal TAG RAM is enabled. If it is then disable it; (It is defective anyway), then try to enable your mobo cache to see if it works properly then. If so then retest w/ Sandra and Everest.
Last edited by Jim on Sat Mar 04, 2006 8:03 pm, edited 2 times in total.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
Post Reply