Page 5 of 8
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 16 Feb 2022 11:13
by frank.lukas
Mega ST4, PAK68/3-030 without Fastram, ST-Low PAL 50Hz, Frontier run from Floppy
40Mhz with 32kB L2 Cache: 3095
50Mhz without 32kB L2 Cache: 1504
... I have only 35ns Tag Ram so 40Mhz is the max. For 50Mhz I need 20ns ones.
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 16 Feb 2022 15:56
by frank.lukas
Atari Mega ST4, EmuTOS 1.1.1, 8MB Alternate Ram, MC68010 8Mhz, boot with nothing in ST-Low PAL 50Hz
frames: 931
Alternate Ram has the same speed as ST Ram 3.7MB. Frontier run from floppy
umbau_klein.jpg
frontier.jpg
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 16 Feb 2022 16:08
by Badwolf
frank.lukas wrote: 16 Feb 2022 15:56
Atari Mega ST4, EmuTOS 1.1.1, 8MB Alternate Ram, MC68010 8Mhz, boot with nothing in ST-Low PAL 50Hz
frames: 933
Alternate Ram has the same speed as ST Ram 3.7MB. Frontier run from floppy
That's an interesting one, Frank. Implies the 010 is approximately the same as doubling the 68k's frequency.
BW
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 16 Feb 2022 16:13
by exxos
Badwolf wrote: 16 Feb 2022 16:08
That's an interesting one, Frank. Implies the 010 is approximately the same as doubling the 68k's frequency.
Odd indeed
https://www.exxosforum.co.uk/forum/viewt ... f=42&t=861
I'll see if I can find a 010 CPU when I get back home later..
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 16 Feb 2022 21:54
by exxos
68010 CPU 889 frames.
68000 CPU 878 frames.
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 17 Feb 2022 08:21
by czietz
My ST, PAL, stock (68000 @ 8 MHz) as reference: 882 frames.
... with Storm ST Alt-RAM: 906 frames.
... with Storm ST Alt-RAM and 68010 @ 8 MHz: 934 frames. <= confirms Frank's measurements.
Similar results are visible in CoreMark, too.
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 17 Feb 2022 12:07
by Badwolf
Thanks, guys. Recorded these on the github.
Also it implies AltRAM measuring the same throughput as STRAM doesn't quite tell the whole tale.
BW.
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 17 Feb 2022 12:29
by frank.lukas
Mega STE 4MB, 16Mhz 68000 + L2 Cache, Blitter off with +6MB Alternate Ram (MagnumSTE)
frames: 1556
The MagnumSTE works not with the Blitter so it´s switch off
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 10 Apr 2022 15:48
by agranlund
Cool benchmark program!
Here's my Atari 520STM (PAL), with 50Mhz TF536.
Running from TT-Ram: 3073
Running from ST-Ram: 1407
My firmware is not stock though. Its got L1 enabled for Rom and ST-Ram access and I believe the stock one only does that for the onboard sdram.
ST-Ram cache misses are not cheap.. but it still makes an noticeable difference for stuff you can't run in fastram without loosing video (ie; framebuffer ending up in fastram)
Re: FrontBench: using Frontier as a benchmark (Discussion thread)
Posted: 10 Apr 2022 20:43
by Elethiomel
agranlund wrote: 10 Apr 2022 15:48
Cool benchmark program!
Here's my Atari 520STM (PAL), with 50Mhz TF536.
Running from TT-Ram: 3073
Running from ST-Ram: 1407
My firmware is not stock though. Its got L1 enabled for Rom and ST-Ram access and I believe the stock one only does that for the onboard sdram.
ST-Ram cache misses are not cheap.. but it still makes an noticeable difference for stuff you can't run in fastram without loosing video (ie; framebuffer ending up in fastram)
Nice! Can you share that firmware please? I'm getting 2940 on a ST536 Rev 5.