Sanity check please...

peteb

Distinguished
Feb 14, 2001
2,584
0
20,780
Hi,

I currently have a PIII 733/133MhzFSB Socket370 CPU running on an ASUS CUV4X mobo.

Overclocking the FSB on this seems very unstable (yes I use 133Mhz RAM) so I was wondering...

If I buy a 100Mhz PIII CPU - and then install it at 133Mhz FSB setup (all components other than CPU running at normal clock speed) this should be more stable right? (providing the CPU is up to it)

If I did this - what is that maximum I'm likely to get out of a PIII?

Can I go for the 850Mhz and clock to 1130Mhz or take the 800 and go for 1064Mhz?

Obviously what ever I do will need some serious cooling so I was going to go with Kyro's recommendation on the Noisecontrol Silverado.

No many will say what a moron, why not go with an AMD and a new mobo? Well, right now I don't want to fully rebuild my system so a chip swap is a nice option - also I'm holding out for the AMD SMP chipsets and then I'll do AMD in style - although I'll needs another Silverado for that... ;o)

Interested on feedback please...

Peter.

-Your Comany Names Could Be Here-
e-mail for application details.
 

Crashman

Polypheme
Former Staff
Most PIII's have dificulty getting past 980MHz. In that respect, the 700 is your best choice, as the 750 goes to 1000@133. Any overclocking will probably require a slight voltage increase.

Suicide is painless...........
 

peteb

Distinguished
Feb 14, 2001
2,584
0
20,780
Hmm - maybe not so worth it then?

I played around last night and managed to get my system running at 150Mhz. I can sustain this buy clcoking the memory back using the 4/3 memory ratio setting (thanks ASUS).

Now my 733 runs at 825 - and I even get better memory bandwidth, according to sandra, by about 5%, even though the memory is running 20Mhz lower than my stock configuration - go figure.

Even using the stock Intel cooler (trying to find someone to get me a Silverado) I can maintain around 42C. I have got the side off with a 20" house fan blowing cold air on the mobo though... ;o)

I might try pushing the voltage up a touch (still 1.65 at present) and see if I can get the bus up a bit further - it won't POST above 150Mhz right now, although the memory should be good for 170 at the multiplier. I suspect that my Geforce Ultra may be complaining about PCI bus speed above 150Mhz though.

Thanks,

I'll let you know how it goes. I doubt a processor forklift is justified, if I could go over 1GHz though that would be cool.... :eek:)

-Your Comany Names Could Be Here-
e-mail for application details.
 
G

Guest

Guest
Your mistakes were in buying the EB chip, and the VIA chipset. As Crashman sez, you won't have any luck trying to get over 1GHz with one of the PIII chips. BX or Solano would have been a better choice, and you'd get 7x140=980MHz from a PIII700E chip (or at the very least, 7x133=933MHz).

I have a 600E running 6x133=800, I have a gaming system based on 7x140=980 (in BX, giving me 92MHz AGP), and an 800E at work running 8x120=960MHz (this one needed 2volts at the core). They all run cool, and the 700E is dirt cheap now. But, as I said, the mistake was buying one of those low internal multiplier EB chips. Sell it if you can, and get the 700E chip.

Clonan
 

peteb

Distinguished
Feb 14, 2001
2,584
0
20,780
Ah - wise words indeed!

Of course the benefit as always is that hindsight is usually 20:20.

Looking back I might have done things differently, but I have had 12 months good stable computing out of the system, and since it is my first self build (actually first home machine I owned after a string of work PCs) I think I did okay. I live in Japan and I was suprised I managed to communicate enough to get components that would actually POST let alone run fast and stable!!!

I am looking for a painless interim step now to sqeeze a few more cycles before I go for a major architecture overhall. I was considering three options...

SMP PIII (expensive, non-scaleable and probably not very productive for anything other than SETI)
AMD (seems like there are a lot of issues shaking out and am waiting for a good selection of 266 boards and AthlonC availability)
AMD SMP (might be a little on the horizon, but I can probably wait - esp. if I can squeeze a little more out of my current system...)

Option 3 is my strategic choice, but it maybe a year+ before a stable mainstream AMD SMP board is out there....

Cheers,

Pete.

-Your Comany Names Could Be Here-
e-mail for application details.
 
G

Guest

Guest
Your memory ratio strategy is an excellent work-around. If you loaded the ASUS probe utility (CD-ROM that came with the board.. or borrow someones's CD), you can keep an eye on the CPU temp to see if the voltage settings increase CPU temp. Your 733 is protected, but it's nice to know how hard your chip is working. The PIII800E that I have running 8x120=960 DID need some extra voltage, but it isn't overheating (shutting down). More cooling would get me to 133FSB, but the system is very stable as is, and it is running great. So, I'll just use it at 960.

I don't think your going to get much past 150FSB ! That is an acheivement ! But extra voltage to the core might improve stability, which is a nice thing to find..
 

peteb

Distinguished
Feb 14, 2001
2,584
0
20,780
Hmm - sounds like a challenge!!!

I'll try going over 150 with the extra voltage...

I've had the Asus probe running the whole time to keep a track of things anf make sure I'm not cooking anywhere obvious.

I have seen some people reduce their AGP from 4x to 2x or even 1x to help POST and boot.

My system was running SETI all night and to my knowledge still is (I'm at work now) so 150 seems VERY stable with the memory clocked way back. It does fail POST 1 in 4, but once up seems solid. Maybe the extra .05V will help the POST and maybe an extra 5Mhz FSB!!!

Cheers,

-Your Comany Names Could Be Here-
e-mail for application details.
 
G

Guest

Guest
Failure to post may not be solved by core voltage increase alone (but might). My P3B-F boards have a system voltage jumper (JP20) that apparently allows switching from default voltage of 3.5 to 3.65V. I forgot to mention that I did this on the PIII800E AS WELL AS the increase to 2.05V core setting (I'm using an ASUS S370-DL slotket convertor which has the core voltage options.. lots of them). If it is marginal voltage for memory, PCI, or AGP support that is causing the posting failures, increasing "system" voltage might support your increased clock settings with more stability.

I also have that 600E chip in a computer at home. It's in another P3B-F. I have a cheap no-name slotket convertor in that system, but jumping to 3.65 allowed me to set 133FSB and go 800MHz with complete stability. This even with TWO SCSI cards: an ISA SCSI II card controlling a UMAX scanner, an internal SCSI insider zip drive, and a 40X Toshiba CDROM, AND a TEKRAM 390F UW SCSI controller, for the 'half height 7200rpm IBM behemoth 30Gb HD (and a second Quantum Viking II 4.5Gb UW 7200rpm drive). All it took there was resetting that system jumper. Maybe your ASUS board has a similar jumper ?
 

peteb

Distinguished
Feb 14, 2001
2,584
0
20,780
Maybe I'll snoop around and look for system voltage jumper - although that sounds a little bit further down the modification path than I was planning on heading...

I'm using a CUV4X which seems pretty friendly for doing a lot of dangerous stuff in BIOS (thank the maker for reset to factory default option).

My system is using onboard UATA/66, USB (audio/game disabled) GeForce Ultra64 Pro, UATA/100Raid and Firewire Combo, SBLive Value and an ISA Datachute PCMCIA slot (for wireless NIC).

All are rather critical to the operation of the system (with the exception of the SB) so I'll see what I can do over the weekend.

With any luck I'll blow the mobo or cpu (or both) and have an excuse to 'upgrade' ;o)

-Your Comany Names Could Be Here-
e-mail for application details.
 
G

Guest

Guest
OOh.. GeForce Ultra64 Pro ?? Well, I wouldn't worry about the AGP channel ! The GeForce chips are overclocking champs. My original GeForce 32DDR (ASUS) is running at 140MHZx2/3=92MHz ! Let's see... 150x1/2 (your CUVX has a 1/2x divider) = 75MHz. Hrumph,, the GPU is not overclocked !!

Heck, the old TNT2 ELSA Erazor in the 600E running 6x133FSB=800MHz is running at 2/3x133=88MHz. No Problem !

Increase the voltage. It'll either post, or not post. If yes, cool, if no.. drop it back and your up and running again ! No upgrade excuse (sorry) !

and remember, AMD means 'Another Meltdown Disaster'
Clonan
 

Crashman

Polypheme
Former Staff
Your processor can actually go much higher, 175FSB should not be too much of a problem at 1.85v (I don't like to go above that, paranoid I guess). But I doubt that the rest of your system can handle it, like the chipset (should go to 160, but maybe higher if your lucky), the cards (especially the PCI cards) etc.

Suicide is painless...........
 

peteb

Distinguished
Feb 14, 2001
2,584
0
20,780
Right - here is where I started (Thanks to SiSoft SANDRA for the benchmarks).

PIII 733/133FSB, 768MB PC133 RAM (3 x 256MB), ASUS CUV4X, Win2K.

What did SANDRA have to say about my CPU and Memory performance?

CPU BENCHMARK
Dhrystone ALU - 1905
Whetstone FPU - 979

CPU MULTI-MEDIA BENCHMARK
Integer SSE - 3962
Floating-Point SSE2 - 4854

MEMORY BENCHMARK
Int ALU/RAM Bandwidth - 309MB/s
Float FPU/RAM Bandwidth - 380MB/s


What did I do? Well,

I tried all sorts of things, but the break for me was removing the original 256MB stick of PC133 I bought with the mobo... Sandra informed me that it was not capable of running CAS2 mode at any speed. The memory description for the other two stick said they'd support CAS2 at 100MHz - which sounded like an opportunity to me...

I tried several combinations and incremental changes in BIOS to get here (downgrading system/memory bus ration to 4/3) but the final steps were...

-----------
150Mhz FSB - CPU and memory (without offending 256MB stick) giving a CPU clock of 825Mhz
Memory set to CAS2
Bump CPU core voltage to 1.85V
This tended to get hot - unfortunately I have not yet found a good source for the Coolermaster ATC200 and Noisecontrol Silverado HSF so I managed to maintain ~ 43C at load (defined by SETI@HOME) using a 20" house fan and no side on my case... I feel bad for asking another poster on the cooling forum if he was serious when he'd said that was what he did....
-----------



What did SANDRA have to say about my new configuration?

CPU BENCHMARK
Dhrystone ALU - 2151 (+13%)
Whetstone FPU - 1104 (+12.5%)

CPU MULTI-MEDIA BENCHMARK
Integer SSE - 4475 (+13%)
Floating-Point SSE2 - 5481 (+13%)

MEMORY BENCHMARK
Int ALU/RAM Bandwidth - 365MB/s (+18%)
Float FPU/RAM Bandwidth - 443MB/s (+16.5%)


CONCLUSIONS - I guess I shouldn't be surprised that a 13% increase in bus speed generates a 13% increase in CPU

performance. I never really expected things to be that logical though - maybe 8 or 9%.

The memory bandwidth was the main gain - and that should help greatly with the old SEIT@HOME throughput...

-* This Space For Rent *-
email for application details
 
G

Guest

Guest
Interesting story... I think with some similar tweaking I could get more out of my systems. The only one that seems near the top (or at the top) is the PIII800E. I have the core at 2.05Volts now, which might be a little more than I need to get to the current 120FSB/40PCI settings. It simply fails to post at 133FSB/33PCI. I could pull the ram and see if if supports the 133MHz setting in the SCSI system, or in the gamer (140FSB). If so, that would confirm that the CPU is the culprit (I think).

There are two additional intermediate settings that I might be able to use. 124FSB/31PCI and 124FSB/41.3 PCI. The latter would push both settings up. This did not work before I increased the core voltage. I haven't tried them yet since the increase. It is really stable at the current settings, and I use the thing constantly for a calculation intensive, HD/RAM/CPU intensive database searching tool. Any down time really cuts into my productivity.

The memory settings are at the default at the moment, and it is cheap CAS3 memory. If I can't go faster without a purchase, I'll probably stay at 960MHz. That is a 20% CPU speed increase, which has to be good. The 40MHZ PCI setting must also be deliverying real time savings. The searches can take 10 to 25 minutes, depending on the search parameters, and the HD is constantly accessed during the searches. These hardware settings are probably buying me a half an hour a day. Maybe more. I can live with that !
 

peteb

Distinguished
Feb 14, 2001
2,584
0
20,780
Indeed - the last few weeks have been intersting as I followed the o/c forum and everone's opinion on how best to do it. I realise now that I was originally very naive about memory and although lucky with my newer RAM to run at CAS2/150 I will probably do things different again.

I have just found a site shipping Corsair PC150 CAS2 256MB sticks for $100 - http://www.spartantech.com - I'm not endorsing them - I haven't used them yet. Might be good for my next system, since I notice they also sell the 133/266 1.333Ghz Athlons.... I see AMD in my future.

I might also go for the PIII750/100 as a replacement to my PIII733/133 - I should be able to get that up to the 1Ghz range without too much trouble!

-* This Space For Rent *-
email for application details
 
G

Guest

Guest
As I mentioned, this gamer/browser toy has a 700E chip happily turning 140MHzFSB/35PCI into a 980MHz system CAS3 spot market PC133RAM !!! After seeing the remarkable performance that I got out of my 700E chip, a friend and colleague of mine decided to attempt a similar strategy.. got a stick of Crucial PC133 CAS2 256Mb, but chose a different MB and CPU: ASUS CUSL and 750E. We both thought that he would make 998MHz with the 133MHz strategy.

I was unfamiliar with the S370 MB, and hadn't learned about the value of core voltage adjustments. We couldn't get the thing past 105FSB with default voltage settings. He took it home, and I haven't convinced him to attempt some tweaking. Tragic, since he bought 1st rate stuff (like the RAM).. CL Live "drive", a Pioneer DVD, and a Yamaha CDRW.

Thing is, I haven't increased the voltage to get this 700E to post 980. Posts on overclockers.com confirm that the 700E is a goldmine. And it is really cheap right now.

The 750 might get to a similar speed, with additional voltage.

Just a word of caution. 700E is cheaper and might go faster.. or just about as fast.
 

peteb

Distinguished
Feb 14, 2001
2,584
0
20,780
Oaky - okay, I'm convinced! :eek:) If I do it I'll get the 700E... *grin*

Lots more to do first. Putting together a new system spec for my tier 1 machine at which point this PIII will head into the living room for my wife to use (although it will be transplanted into a much more pretty case.)

Later,

Pete

-* This Space For Rent *-
email for application details