OverClocking the DFI K6BV3+/66

Discuss software and how to tweak more performance out of your system.
Post Reply
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

OverClocking the DFI K6BV3+/66

Post by Jim »

Yeah, I finally managed to do it. Not much, just 618 MHz; but at least now I know how to do it; so now the problem is to get it stable up there.

For those not familiar w/ the DFI K6BV3+/66, it is one of those boards that has no OC settings on the board or in the bios. You have to use software to OC it. I am using CPU Cool to OC mine. The catch to that was that CPU Cool lists the DFI K6XV3+/66; but not the K6BV3+/66 as being among the boards supported. My mistake was to set it at K6XV3+/66, thinking they were pretty much the same. (WRONG). The thing to do for the K6BV3+/66 is select DFI as the manufacturer, but do not select any particular board. Instead go to the next screen and select the PLL. That works.
Last edited by Jim on Mon Oct 10, 2005 7:22 am, 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
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

RE: OverClocking the DFI K6BV3+/66

Post by Jim »

Well actually it was 619.2 MHz @ 2.0v. It ran Everest alright; but Sandra crashed it instantly. Raised the voltage to 2.05, still ran Everest ok; and still crashed under Sandra. Lowered the multiplier to 5.5, still @ 2.05v; and was able to run Sandra; so I think it is a voltage issue. Will try going back to 2.1v. @ 6x103.2 = 619.2 MHz to see how that works before attempting to go up to 112 MHz FSB.

Edit: K, been there done that : K6-3+ 450 ACZ @ 2.1v. on DFI K6BV3+/66 w/ 1 Meg Cache @ 6x103.2 MHz.

Everest :
Memory Read = 313 MB/s
Memory Write = 149 MB/s
Memory Latency = 208.4 ns

Sandra :
Memory Bandwidth :
Int MMX = 205, 211, 216, 215.
Float FPU = 208, 212, 214, 214.
Arithmetic :
Dhrystone MIPS = 1458
Whetstone MFLOPS = 770
Multimedia :
Integer MMX = 3503
Float 3D NOW! = 4152

The Cache and Memory Bandwidth test failed @ 2.1v. It looks like I will have to try to "burn in" this processor before going further.
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
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

RE: OverClocking the DFI K6BV3+/66

Post by Jim »

Did a test @ 112x5.5 = 616 MHz. Everest crashed it. "Burn in" in progress. Hope I am doing it right. Have set the FSB @ 3x100, voltage @ 2.4v. Have Sandra running the burn in module w/ everything but "removable storage" & "file system" modules enabled. Am also running "WinAmp 5", as well as various third programs. i.e. Spyware check, Antivirus check, ATI "File player" etc. Should be finished by Monday night. See then if it makes any difference.
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
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

RE: OverClocking the DFI K6BV3+/66

Post by Jim »

The latest is this: 1) My ram is unable to support 2,2,2,5 timings @ 112 FSB. It will @ 103.2 FSB but not @ 112. 2) The L3 cache on my DFI K6BV3+/66 1 Meg board must be disabled to run 112 FSB. This means I take a substantial memory performance hit going from 103 to 112. That is unfortunate because the 5.5x multiplier works better for memory than the 6x does; and I had hoped to set this machine @ 5.5x112. I can do it; but only w/ slackened ram timings and L3 cache disabled, = worse than 6x103.

Net result will have to get better Ram, (i.e. PC133 2,2,2,5.), before I go up to 112 FSB. Apart from that, the burn in was a success, enabling the board to run stable @ 6x103.2, completeing all tests so far successfuly. Also would note the processor will now clock : 450MHz @ 1.3v., 500MHz @ 1.45v., 550MHz @ 1.55v., and will at least boot into windows @ 600MHz @ 1.85v. Test results were as follows:

K6-3+450ACZ @ 6x103.2 on DFI K6BV3+/66 w/ 1 Meg cache & 512 Meg Hynix SDRAM.

Everest :
Memory Read = 312 MB/s
Memory Write = 149 MB/s
Memory Latency = 208.0 ns

Sandra :
Arithmetic 1460 MIPS ; 771 MFLOPS.
Multimedia Int MMX 3507 it/s Float 3D Now! 4158 it/s
Memory (Test run several times because of rising numbers)
Int MMX : 206; 212; 215; 216
Float FPU : 209; 212; 214; 214
Cache & Memory (Total MB) = 10263; 10255.
All tests completed @ 2.1v.
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
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

suggestion for benchmarking

Post by DonPedro »

hello jim!

always appreciate to read from you ....

but I would like to make a suggestion: while everest suits well for a quick shot on memory performance, it does not say a lot about how that translates into performance-gains/losses in real-world applications. the same applies to the often referred to tool cpumark and sandra.

so in order to see the what appliactions benefit or loose because of different fsb, multiplier settings or whatever other tweaks, I suggest you check out "osmark"

http://www.vanshardware.com/news/2004/0 ... 040422.htm

beware that an "official" osmark-run takes about one and half an hour to complete (it runs all tests three times and then takes the best run as final result; on the other hand its quite interesting to watch some of the tasks being performed). you will get scores on all the various single tasks and also an overal osmark-mark.
you can save your results in an inbuild database, export and import results (good for sharing results). the people behind that benchmark suite also have a forum running for whatever one wants to discuss:

http://www.flickerdown.com/osmark/index.php

another less time-cosuming test-suite, but also not so much real-world guided tool would be hot cpu tester, a burn in and stability tool which also offers a benchmark test-suite.

http://majorgeeks.com/download273.html

You will be surprised of when and to what extent cpumark, sandra and everest scores translate into meaningful real world applications.

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

RE: suggestion for benchmarking

Post by Jim »

Peter, I have downloaded OSMark; but frankly it looks to be somewhat over my head. I will run some tests w/ it to generate some comparison numbers, for some of the things I have tried; but it is so slow compared to Everest, that it is a major disadvantage. Everest may not relate to particular results very well; but it does give you a fast way of seeing : Better ? or Worse?
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
tazwegion
Veteran K6'er
Posts: 349
Joined: Sat Jun 21, 2003 11:13 am
Location: Victoria, Australia

RE: suggestion for benchmarking

Post by tazwegion »

* Prime95 is a wonderful benchmark for testing system stability similar to Sandra's stress testing suite ;)

* CPUmark99 is a good means of measuring system perfomance (available from the downloads section @ my site :P)

* 3Dmark99 (onwards) is an excellent graphics intensive benchmark suite also

BTW when are you gonna' start posting some pic's for us Jim? ;)
Image
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

RE: suggestion for benchmarking

Post by Jim »

Taz, that is something I would like to do because I am (sin) kind of proud of the mods I made to this machine. That said, It is low priority at the moment, because I am in the middle of a 40 gig data rercovery, and trying to get this thing stable up at 619 MHz. Looks like I may have to go to 2.2v, or alternately change my ram to do it. Sandra & Everest are not the most challenging programs out there. Don Pedro's test suite crashed me 3 times just opening up the picture files to take a look at them. (BTW Don Pedro it came w/ 1 of the picture files corrupted so it won't complete properly. I could try redownloading to see if that solves the problem.) EDIT: It did solve the problem. I suspect it was an instability issue. Gonna have to try 2.2v. and see if that clears it up. If not will have to go back to 600MHz till I get new ram.

I have to dig out my polaroid, take some pics, scan them, (probably buy some film too, don't have enough, to show all the changes I 've made in this sucker.) (2 extra internal bays, ducted fan setup, home made plug receptacles for wiring, dust filter setup, clip in almost everything etc.) Then having done all that, I will be back to ask you WTF is a "URL"?

BTW Taz, how do you keep your post total so low? I would like to get mine back down to about 60 or so to better reflect my (knowledge) standing in the K6Plus community.

Also should say, Taz, take a look at this test suite of Peter's. Wow!! Numbers, Numbers, Numbers!! -- What they mean? -- Who knows? - Not me that's for sure; but it sure does give you a detailed picture.
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
tazwegion
Veteran K6'er
Posts: 349
Joined: Sat Jun 21, 2003 11:13 am
Location: Victoria, Australia

RE: suggestion for benchmarking

Post by tazwegion »

LMAO I must've answered that question half asleep :lol: I actually meant the Image tags :P

I assume that URL equates to User Referal Link or similar ;)

As far as the appropriate coding to post pictures here is an example...


http://imageshack.us/img/imageshack.png (site address of host) / (filename.filetype) ;)

So posting that picture file would look like this...

Code: Select all

[img]http://imageshack.us/img/imageshack.png[/img]
The end result would be...

Image

I look forward to seeing these SS7 modded PC pic's soon! :banana and as far as the low post count... I simply post OR respond to whatever takes my fancy :lol: plus I'm a member of @ least 5 other forums :twisted:

BTW the picture is the imageshack.us logo ;)
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

regarding osmark

Post by DonPedro »

.
@jim

well, you're right, it is a disadvantage that it takes so long to run the "official"-mark. but then the medal has two sides: by running real-world applications (mp3-conversion, encryption of files, packing of files, working on some long text-file, altering pictures in whatever way) you finally know what the tweak is worth. and having the system perform these tasks takes time - as in real life.

of course, you're right again in that for a first look whether a tweak produces any change and in what direction, everest and such is really great.

but if you want to save time, then check out those applications in osmark that are sensitive to e.g. memory tweaks. an approach would be by reversing memory tweaks to its default state within bios, no use of wpcredit, running osmark, then setting all memory relevant parts to top performance and have osmark run again. compare the individual results within the test-suite and you know which single tests are memory-settings sensitive. in future, run them alone when you want to test new memory tweaks, there is no need to run the official mark everytime. this will save you A LOT OF TIME! :)

btw, did you realize that within the result viewer, you can zoom in the results by clicking on the respective part in the result bar?

I also found it useful to delete the results which came preinstalled. because the graph of results takes the best performer and relates all other results graphically to it. this leads to almost indiscernible small graphics of our k6-systems compared with athlon 64. (you can "save" the athlon etc results by exporting it prior to deletion).


@taz:

I don't think that cpumark is a very useful tool to measure system performance. the tool responds heavily on on-die cache of the cpu. just run it on a k6-2+ and on a k6-3+ (all other things keeping equal) and it will show a dramatic increase in performance. this is actually not true. throw both cpus at osmark and the picture is corrected by reality. mmmh, I think I will do that cpumark/osmark comparison and post the results here ......
User avatar
tazwegion
Veteran K6'er
Posts: 349
Joined: Sat Jun 21, 2003 11:13 am
Location: Victoria, Australia

RE: regarding osmark

Post by tazwegion »

While CPUmark99 is primarily measuring processor performance, it also reflects m/board & chipset interaction... and I never use it stand-alone per say ;)

For example an SiS530 SS7 + K62-500 pulled 18-21 points, the same CPU & memory on a VIA MVP3 platform yeilded 28 points etc.

Hmmmm... OSmark? I'll have to look into this one :D
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

RE: regarding osmark

Post by Jim »

Do that!!! You will get your mind blown. Not what you expect at all. Sort of like comparing a symphony orchestra to a guy w/ a harmonica.
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 »

Something else to consider is when you actually run the test...and what is running/has been run prior to the time of the test.

I normally run all testing from a clean boot after the system has "stabilized".

Don't forget that startup items (such as your anti-virus) can affect your results...just as adding OS patches, etc... so it is a good idea to keep track of that as well...when you think you have found the *magic*... :wink:

What DonPedro says has merit...just because a system benchmarks well doesn't mean it will perform the same way in real life applications... :idea:
User avatar
tazwegion
Veteran K6'er
Posts: 349
Joined: Sat Jun 21, 2003 11:13 am
Location: Victoria, Australia

Post by tazwegion »

Of that I am well aware, all benchmarks are synthetic, real world applications is where it counts... for example I've got a 32Mb Diamond Viper 2 (z200 chipset) during 3Dmark2000, it absolutely flogs every other AGP card in my humble collection... with the exception of the GF2 & R9600XT, however playing V-rally2 it loses resolution & pixelates, becoming visually out-performed by both of the 3Dfx Banshee 16Mb cards PCI / AGP & 32Mb TnT2pro (modded) which do not exibit such flawed graphics ;)

Ah that's life in the FAST lane :banana
Post Reply