CoreMark and LINPACK benchmarks for Atari

All about the serious stuff.

Moderators: Mug UK, Zorro 2, Moderator Team

User avatar
Arne
Atari Super Hero
Atari Super Hero
Posts: 746
Joined: Thu Nov 01, 2007 10:01 am

Re: CoreMark and LINPACK benchmarks for Atari

Post by Arne »

DarkLord wrote: Tue May 18, 2021 8:23 am And it appears that your 40mhz Pak 68/3 is also beaten by the 32mhz '030 Hypercache board.
Not in absolute numbers:
HC030@32MHz = 11.26
PAK68/3-030@40MHz = 11.63

But when it comes to CoreMarks per Mhz then the HC030 is faster.
Image
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Just curious, does that seem right to you, considering the
Pak board is running 8mhz faster?

Thanks.
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
User avatar
Arne
Atari Super Hero
Atari Super Hero
Posts: 746
Joined: Thu Nov 01, 2007 10:01 am

Re: CoreMark and LINPACK benchmarks for Atari

Post by Arne »

You are right. At first sight it looks suspicious. PAK does have a L2 of double size and more clock speed. Only difference I see is the split L2 on the HC030 which might make a difference for the CoreMark algorithm.
My guess is that the CoreMark code is overwritten by data more often in the unified L2 of the PAK than it is in the split L2 design of the HC030. But this is like reading tea leaves.

But I wonder why your PAK68/3 030 is slower than mine at the same clock speed!
Image
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Good question. When I received my board (off Ebay) it was damaged. Something the
seller did not inform me about until I received it. Eventually, I was able to get it working
but I've never been able to get it successfully running at 50mhz. It will work there, but
it's unstable. I'll get odd, random crashes from software at various times. It is rock
solid at 40mhz though, literally running for hours and hours with no problems.

I'm wondering if something from the damage is holding it back a bit.

PS Just to check, did you run coremark with NVDI active? I did not.
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
czietz
Hardware Guru
Hardware Guru
Posts: 1539
Joined: Tue May 24, 2016 6:47 pm

Re: CoreMark and LINPACK benchmarks for Atari

Post by czietz »

DarkLord wrote: Tue May 18, 2021 9:04 pm PS Just to check, did you run coremark with NVDI active? I did not.
CoreMark measures CPU and RAM (and Cache) performance. NVDI has no effect on it.
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Understood. Thanks! :)
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
czietz
Hardware Guru
Hardware Guru
Posts: 1539
Joined: Tue May 24, 2016 6:47 pm

Re: CoreMark and LINPACK benchmarks for Atari

Post by czietz »

DarkLord wrote: Mon May 17, 2021 3:22 pm Okay, thanks. Not sure about Linpack then. I can set it to something like "10" and it runs,
but seems to take forever. Anything near the 200 setting, 2 bomb crash (not fatal, back
to desktop).
I have confirmed with another PAK68/3 owner that Linpack runs correctly on that card (provided a FPU is installed, of course). Thus, something must be strange with your setup. Or maybe the program file was damaged/corrupted during transfer.

Unfortunately, TOS's 2 bombs (= bus error) is too little information to narrow it down. (I wish Atari had opted to print a full exception trace much like EmuTOS does.)
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Okay thanks. It is probably something to do with my (originally) damaged
setup.

Having a 68882 vs the 688881 shouldn't make a difference, right?
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
czietz
Hardware Guru
Hardware Guru
Posts: 1539
Joined: Tue May 24, 2016 6:47 pm

Re: CoreMark and LINPACK benchmarks for Atari

Post by czietz »

DarkLord wrote: Wed May 19, 2021 5:25 pm Having a 68882 vs the 688881 shouldn't make a difference, right?
The 68882 will be faster, but Linpack runs on both, as can be seen from the result table: https://github.com/czietz/linpack-atari/

EDIT: Would be interesting to see what Nembench says for your PAK, particularly in terms of floating point performance: https://milan.kovac.cc/atari/software/? ... MBENCH.FTP
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

I'll try it out and report back later...
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Okay, so I downloaded and ran NEMBENCH.LZH.

Here's a screenshot of the results:

P1010002.JPG
You do not have the required permissions to view the files attached to this post.
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
czietz
Hardware Guru
Hardware Guru
Posts: 1539
Joined: Tue May 24, 2016 6:47 pm

Re: CoreMark and LINPACK benchmarks for Atari

Post by czietz »

DarkLord wrote: Thu May 20, 2021 5:04 am Here's a screenshot of the results:
I don't have a PAK to compare to, but nothing looks out of the ordinary. In particular, you can see that the FPU (where it says "float") is running at full speed. The 250% are to be understood in relation to the Falcon @ 16 MHz.
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Well, that's good.

Thanks. :)
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
czietz
Hardware Guru
Hardware Guru
Posts: 1539
Joined: Tue May 24, 2016 6:47 pm

Re: CoreMark and LINPACK benchmarks for Atari

Post by czietz »

That being said, NEMBENCH only tests the speed the FPU operates at, but not if it does the correct calculatations. In contrast, LINPACK relies on correct results, otherwise it can crash or take forever.

There's a FPU tester: https://www.atari-forum.com/viewtopic.p ... 85#p410785
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Oh, that's Doug Little's tester right?

I'll download it and give it a shot.

Thanks.
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Hmm, not sure about what's going on here.

I ran FPUTEST.TOS from that zip file and got about half errors. (failed)

I ran FPUREC.TOS and it passed with flying colors - no failures at all.

But according to the README file, FPUREC.TOS is theoretically for '040 and '060 machines.

Or am I reading this wrong?

Thanks.
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
czietz
Hardware Guru
Hardware Guru
Posts: 1539
Joined: Tue May 24, 2016 6:47 pm

Re: CoreMark and LINPACK benchmarks for Atari

Post by czietz »

DarkLord wrote: Sat May 22, 2021 11:06 pm I ran FPUTEST.TOS from that zip file and got about half errors. (failed)
I ran FPUREC.TOS and it passed with flying colors - no failures at all.
But according to the README file, FPUREC.TOS is theoretically for '040 and '060 machines.
Hm, no. FPUREC.TOS records the results of a known good FPU into a .DAT file that can then be used to test a suspect FPU (of the same model) by FPUTEST.TOS. But the EXPECT.DAT in the archive is already correct for 68881/68882 FPUs. Therefore, if your FPU fails FPUTEST.TOS (with the provided EXPECT.DAT), I'm afraid your FPU is faulty. :(
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Wow. That sucks. Yet it never crashes or gives errors on any other bench marking software
or any software that I've thrown at it that uses the FPU.

Go figure. :(
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

For whatever reason, Linpack finally decided to play nice and not crash:

(but I'm not happy with the results considering a stock TT with a 32mhz FPU
beat my scores, according to the results chart). :(


P1010005.JPG
You do not have the required permissions to view the files attached to this post.
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Hate it when things are just crazy...

Finally got around to installing a 68881 into my Mega STe. I bought the pack from
Best Electronics, brand new 68881@16mhz and the GAL along with it. Installation
went fine, my Mega STe's MB had the jumper so I didn't have to add a resistor to
the bottom of the board.

Sysinfo says I have an FPU. Gembench says I have an FPU but won't measure it because
it's an "I+O based FPU". Coremark ran fine (but I think an FPU with it is irrelevant)...

FPUTEST crashes out hard with 6-7 bombs - won't even begin to test anything... :(
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
User avatar
DarkLord
Fuji Shaped Bastard
Fuji Shaped Bastard
Posts: 4720
Joined: Mon Aug 16, 2004 12:06 pm
Location: Prestonsburg, KY - USA
Contact:

Re: CoreMark and LINPACK benchmarks for Atari

Post by DarkLord »

Guys, I must be reading this wrong:

https://www.atari-forum.com/viewtopic.p ... st#p304768

DML's first line says that it's for TT's and Falcon's?
Welcome To DarkForce! http://www.darkforce.org "The Fuji Lives.!"
Atari SW/HW based BBS - Telnet:darkforce-bbs.dyndns.org 520
Post Reply

Return to “Applications”