Who has the lowest benchmarks? - Page 4

QuintLeo · 2003-07-19, 17:58

Quote:

Originally Posted by Xyzzy

IIRC, GIMPS requires at least a 386, so technically a 16MHz 80386 should be the slowest processor possible, unless you want to count Macs... shock lol

386SX would be slower - and they did have a 387SX option.

*wince*

I remember when I thought that my 80286-10 was a blazingly fast machine. Nowadays I wouldn't even THINK of using one, except mayhaps for a dedicated BBS box.

Ooohh - 384K just finished. Best time 85866.844 ms

I think we have a "legitimite" winner here.... 8-)

(edit) 448k Best time 179666.610 ms

QuintLeo · 2003-07-19, 18:00

Quote:

Originally Posted by nomadicus

IBM ThinkPad
Intel 486 processor
CPU speed: 100.00 MHz
L1 cache size: unknown
L2 cache size: unknown
L1 cache line size: unknown
L2 cache line size: unknown
Prime95 version 23.4, RdtscTiming=1
Best time for 384K FFT length: 511272.479 ms.
.
.
.
Best time for 2048K FFT length: 3004013.919 ms.

Hmmm. Those numbers still don't look right. I wonder how much time your laptop spent in "power saving, down-clock the CPU" mode?

Xyzzy · 2003-07-19, 19:46

Quote:

Originally Posted by QuintLeo

I remember when I thought that my 80286-10 was a blazingly fast machine. Nowadays I wouldn't even THINK of using one, except mayhaps for a dedicated BBS box.

Up until last year I used my PS/2 Model 30 (8086) as a dedicated serial console...

delta_t · 2003-07-19, 20:49

hey nomadicus,
perhaps you are right, it may have misread the 486. if in fact it is really a true 33mhz laptop than it would be pretty slow, but i'm not sure if it would be that slow. quintleo just started a 386dx so we should see how that compares. what do you think we'll hear from him in 3 weeks?

QuintLeo · 2003-07-20, 02:38

Sooner - I'm limited to putting 16 megs total in that machine.

1) 4 sockets, 72-pin SIMMS

2) Won't work with 8 meg SIMMS period - appears to not support double-sided RAM sticks at all.

3) Won't work with the only pair of 16-Meg SIMMS I have - they're non-parity, the MB requires parity.

The results I did get

CPUID not supported - 486 CPU assumed (386DX/387DX)
CPU speed 100.00 MHz (33/25 actually. I dunno where that 100Mhz came from).

I'm guessing that it assumes a 486DX/100 if the CPU doesn't support CPUID - which would explain the mis-IDed 486 laptop earlier in the thread (CPUID was initially added in the Pentium, then retrofited to some of the later clock-multiplier 486 class chips).

L1 cache size unknown
L2 cache size unknown
L1 cache line size unknown
L2 cache line size unknown
Prime95 version 23.5, RdtscTiming=1

Best time for 384K FFT length 85866.844 ms.
Best time for 448K FFT length 179666.610 ms.
Best time for 512K FFT length 195583.491 ms.
Best time for 640K FFT length 262386.336 ms.
Best time for 768K FFT length 314203.225 ms.
Best time for 896K FFT length 380778.592 ms.

I got a Segmentation Fault at 1024K FFT - I dunno if it hit a bad RAM stick, or just ran out of RAM period. I tried swapping some different RAM in and am trying it again as a double-check.

Given the trend in timings, I'd guess a 1792K FFT would take somewhere in the ballpark of 760 seconds per iteration. *eep*

mprime 23.5 running under a fairly minimal Slackware 8.1 install (Slack 9 is compiled for 486 or better, and I didn't feel like taking the time to recompile) using 2.4.18 kernel.

Update 21July - The re-run with additional RAM produced essentially identical results. I'm going to have to do more playing with the RAM, it appears - I've been able to benchmark other systems under LINUX with 16 Megs RAM in them, I thought....

nomadicus · 2003-07-21, 17:45

Quote:

Originally Posted by QuintLeo

Hmmm. Those numbers still don't look right. I wonder how much time your laptop spent in "power saving, down-clock the CPU" mode?

I don't think a 33MHz 486 had power saving features???

nomadicus · 2003-07-21, 17:49

Quote:

Originally Posted by delta_t

hey nomadicus,
perhaps you are right, it may have misread the 486. if in fact it is really a true 33mhz laptop than it would be pretty slow, but i'm not sure if it would be that slow. quintleo just started a 386dx so we should see how that compares. what do you think we'll hear from him in 3 weeks?

Yeah, I know. My results look too pesimistic. I'll turn it back on next chance I get and watch the post more closely to see if I missed anything relevant.
3 weeks? Naw . . . 2 weeks, 6 days and 19 hours. 8)

QuintLeo · 2003-07-22, 13:46

Almost ANY laptop has power-saving features - all the way back to the GRID 8088 thingie.

Doesn't have to have CPU clock slowdown IN the CPU - that can easily be implimented in the BIOS with one of the many common programmable clock generator chips.

In any event, *some* 486 class chips did have power-saving features. I *think* a few of the last 386-class chips did too, but not as certain about that.

Paulie · 2003-07-24, 21:19

Quote:

Originally Posted by QuintLeo

Almost ANY laptop has power-saving features - all the way back to the GRID 8088 thingie.

Wasn't that the one with the bubble memory??? I knew a guy who had a Grid with bubble memory. :)

Xyzzy · 2003-07-25, 04:37

Quote:

Originally Posted by Paulie

Wasn't that the one with the bubble memory??? I knew a guy who had a Grid with bubble memory. :)

http://www.maineantiquedigest.com/articles/l030896.htm

guido72 · 2003-07-28, 18:30

Quote:

Originally Posted by Xyzzy

IIRC, GIMPS requires at least a 386, so technically a 16MHz 80386 should be the slowest processor possible, unless you want to count Macs...

:D

Hi Mike and Hi all!
As you Know the Sun of July may be dangerous for our CPU and for our Brains...

I've tryed to run Prime95 Ver.16 on my very old 386 system with just 8 Mb of Ram. No way to run latest versons of George's Program, so no Benchies at all... Just a screenshot...
http://it.geocities.com/guy1972_it/Prime95_on_386.jpg

The only good thing: even running Prime95, CPU doesn't need any Heatsink....

Regards from Italy

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Lowest Unknown Prime	GuyMacon	PrimeNet	6	2011-05-07 03:20
Benchmarks	MurrayInfoSys	Information & Answers	3	2011-04-14 17:10
LLR benchmarks	Oddball	No Prime Left Behind	11	2010-08-06 21:39
lowest 10M exponents available - pick your choice	tha	Lone Mersenne Hunters	116	2008-11-23 13:03
What is the Lowest Rank you can have	crash893	Data	7	2006-01-26 05:26

2003-07-19, 20:49	#37
delta_t Nov 2002 Anchorage, AK 357₁₀ Posts	hey nomadicus, perhaps you are right, it may have misread the 486. if in fact it is really a true 33mhz laptop than it would be pretty slow, but i'm not sure if it would be that slow. quintleo just started a 386dx so we should see how that compares. what do you think we'll hear from him in 3 weeks?

2003-07-20, 02:38	#38
QuintLeo Oct 2002 Lost in the hills of Iowa 2⁶·7 Posts	Sooner - I'm limited to putting 16 megs total in that machine. 1) 4 sockets, 72-pin SIMMS 2) Won't work with 8 meg SIMMS period - appears to not support double-sided RAM sticks at all. 3) Won't work with the only pair of 16-Meg SIMMS I have - they're non-parity, the MB requires parity. The results I did get CPUID not supported - 486 CPU assumed (386DX/387DX) CPU speed 100.00 MHz (33/25 actually. I dunno where that 100Mhz came from). I'm guessing that it assumes a 486DX/100 if the CPU doesn't support CPUID - which would explain the mis-IDed 486 laptop earlier in the thread (CPUID was initially added in the Pentium, then retrofited to some of the later clock-multiplier 486 class chips). L1 cache size unknown L2 cache size unknown L1 cache line size unknown L2 cache line size unknown Prime95 version 23.5, RdtscTiming=1 Best time for 384K FFT length 85866.844 ms. Best time for 448K FFT length 179666.610 ms. Best time for 512K FFT length 195583.491 ms. Best time for 640K FFT length 262386.336 ms. Best time for 768K FFT length 314203.225 ms. Best time for 896K FFT length 380778.592 ms. I got a Segmentation Fault at 1024K FFT - I dunno if it hit a bad RAM stick, or just ran out of RAM period. I tried swapping some different RAM in and am trying it again as a double-check. Given the trend in timings, I'd guess a 1792K FFT would take somewhere in the ballpark of 760 seconds per iteration. eep mprime 23.5 running under a fairly minimal Slackware 8.1 install (Slack 9 is compiled for 486 or better, and I didn't feel like taking the time to recompile) using 2.4.18 kernel. Update 21July - The re-run with additional RAM produced essentially identical results. I'm going to have to do more playing with the RAM, it appears - I've been able to benchmark other systems under LINUX with 16 Megs RAM in them, I thought....

2003-07-22, 13:46	#41
QuintLeo Oct 2002 Lost in the hills of Iowa 700₈ Posts	Almost ANY laptop has power-saving features - all the way back to the GRID 8088 thingie. Doesn't have to have CPU clock slowdown IN the CPU - that can easily be implimented in the BIOS with one of the many common programmable clock generator chips. In any event, some 486 class chips did have power-saving features. I think a few of the last 386-class chips did too, but not as certain about that.