max memory with a K6-X+ series?

Discuss software and how to tweak more performance out of your system.
Post Reply
fla56
Newbie K6'er
Posts: 3
Joined: Tue Jul 30, 2002 11:25 am

max memory with a K6-X+ series?

Post by fla56 »

i remember learning when i first got my system (MVP3) that i couldn't install more than 256meg of memory without screwing up L2/L3 caching.<br><br>anyone know if this applies to the mobile 'plus' series of chips, i heard somewhere (not for sure) that they've got some different kind of addressing than means it's not a problem, maybe? <p></p><i></i>
io333
K6'er Elite
Posts: 888
Joined: Fri Jun 08, 2001 11:11 am

don't worry

Post by io333 »

it does not apply to the plus (+) series chips. <p></p><i></i>
User avatar
AZK6er
Veteran K6'er
Posts: 303
Joined: Tue Jun 05, 2001 2:27 am

Re: don't worry

Post by AZK6er »

OK then, I'm running a Tyan s1590s at100 board w/K6III+ 450 OC'd to 600 with 256M of Crucial (2x128M sticks) and my board has an open slot for another 128M stick, taking the total to 384M. I always wondered if I could go all the way to board max (384M) without a performance hit. I haven't as of yet due to re-working a friends K6-2 400 that he had 512M of CAS-2 installed and ran like a PC-XT. I took him down to 192M and BAM!!, now he can play ALL his games. That made him just happier than a lark. Thats why I haven't gone to 384 yet. But you say I can? <p></p><i></i>
bronx69
Junior K6'er
Posts: 42
Joined: Thu Jun 07, 2001 6:57 am

and your max is

Post by bronx69 »

it depends on the amout of L2 cache you have.<br>if you have 2MB of L2 cache and you using a K6-III+<br>your max amout of ram is 576 and even at that amout expect to see a proformance hit. 512 is optimal for most users and strongly recomended by me. <br>your system will run fastest with one stick of 256<br>vs two @ 512. however the hit is not much an the ram is welcome. <br><br>you can even go up to 768 however i do not recomend it. you can however do it with your L2 cache disabled and go as high as 1g but this defeats the whole purpose a having a K6-3 with a 2mb L2 cache system.<br> <br> To some this is not a bad idea because you may be able to exceed 600 mhz and step into the 700mhz range. but please note that this far exceeds the MVP3 chipset 500mhz spec so chipset failure is a reallity. <br><br><br> <p></p><i></i>
io333
K6'er Elite
Posts: 888
Joined: Fri Jun 08, 2001 11:11 am

i differ

Post by io333 »

i ran many a benchmark with the + chips, with l2 cache on and off, with lots of ram and little ram, and believe me, in the real world, any benchmark can go either up or down without any direct correlation as to whether the l2 is on or off. <p></p><i></i>
760MPX
Newbie K6'er
Posts: 20
Joined: Tue Jul 10, 2001 6:46 am

Re: max memory with a K6-X+ series?

Post by 760MPX »

IMHO, write allocation (or the lack of) has more impact on performance than the amount of L2/L3 cache for a K6-III or K6-X+ CPU on a mainboard with the MVP3 chipset. MVP3 supports write allocation only when you have <= 384MB (not 100% sure; maybe 256MB; please check). If you have more RAM than that, it simply does NOT do any write allocation. Then if you don't use a utility like K6Speed to enable write allocation yourself, you'll get a performance hit much bigger than if you disable the L2/L3 cache. This has been verified by friends of mine who have the DFI K6BV3+ with 2MB cache.<br> <p></p><i></i>
jwhickman
Junior K6'er
Posts: 46
Joined: Fri Nov 01, 2002 8:55 am

K6-[2/3]+ cacheable limits, tag ram, and L2

Post by jwhickman »

I researched the he!! out of this a couple months ago when I upgraded my K6-2+ 450 from an Asus TX97XE to a P5A-B AladdinV revE (later rev's (F and/or G?) had problems w/ tag ram and k6-*+ cpus; I think F disabled the on-mobo L2 and the G fixed the tag ram problem, but caused major performance problems). Anyway, I found the following...<br> 1. The k6-3 and the + cpu's are not constrained by the -same- chipset, tag ram, and on-mobo L2 issues; I first verified that performance decreased using a K6-2 (not '+') going from 64 to 128mb on the TX97XE and then also going from 128 to 256mb (may have been from 256 to 320, but I don't think so) on the P5A-B. 64/128(or 256?) are the respective known cacheable limits of the 430TX/AladdinV chipsets on the TX97XE/P5A-B boards, all respectively. Then I put in my k6-2+ and verified that performed the same test INCREASED performance, thus showing that the + cpu doesn't have the same cacheable limit problem. <br> 2. The cacheable limit of the K6-2+ is 512MB and of the K6-3+ is 1024MB, I believe. This is based on research which was VERY hard to track down; one good source I remember (out of a couple) was at...<br><!--EZCODE AUTOLINK START--><a href="http://www17.tomshardware.com/howto/00q ... <!--EZCODE AUTOLINK END--> and it seemed like anandtech and/or geek.com had some more info.<br> 3. I had to disable the on-mobo cache (now L3) because I couldn't run stable over 100mhz fsb with it. With it disabled I could go up to 125mhz fsb. I found that while non-memory intensive benchmarks went down a little, memory benchmarks went WAY up w/ it disabled, I imagine since it running at PCI speed caused asychronousity issues w/ the cpu and on-die L2. The increased FSB far outweighed the loss of the L3, though.<br> 4. I used k6speed to enable write allocation w/ my TX97XE (setup to run automatically at startup, enable it, then exit), but when I got the P5A-B, I found Asus had just released a new BIOS (around Aug. 02) that did some power-mgmt and win2k stuff but I also found that it now recognized the + cpu, indicated the correct speed, and correctly enabled write allocation. How's that for product support!<br>Well, that all...<br>Jeremy <p></p><i></i>
cooliscool123
Junior K6'er
Posts: 43
Joined: Fri Nov 01, 2002 10:29 pm

Well...

Post by cooliscool123 »

I've had 768MB PC-133 @ 110mhz in this system. (512K L2, Via MVP4 Chipset), and know a guy who has a K6-2 550 (5.5x100) which uses 1024MB Pc-133. Works great. <p></p><i>Edited by: <A HREF=http://pub70.ezboard.com/bk6plus67153.s ... cool123</A> at: 11/9/02 12:19:45 am<br></i>
Post Reply