SAM440 - R9250 vs HD 7750 benchmarks

Hello benchmarks maniacs

After success with a HD7750, I have made several tests, Here we go...

System Information:
OS: AmigaOS Release 4.1 - Final Edition Update 2
Motherboard: Sam440EP
CPU: AMCC PPC440EP, 1.3 @ 800 MHz
L1 Cache Size: 32768, L2 Cache Size: 0, L3 Cache Size: 0
Total RAM: 1 GiB
External Bus (FSB) Speed: 133.333 MHz
Exec.library version: 54.30
Graphics.library version: 54.248
Intuition.library version: 54.28
Picasso96API.library version: 54.18
Rtg.library version: 54.90



Board name: Radeon R9250......................................Board name: Radeon HD 7700 Series
VRAM: 127.984 MiB...............................................VRAM: 1008 MiB
Display mode: 1920x1200@60 (32 bpp)....................Display mode: 1920x1200@60 (32 bpp)

Copy from RAM to VRM:

copy32: 56.425 MiB/s (took 0.153558 seconds)...........copy32: 39.349 MiB/s (took 0.220196 seconds)
copy64: 57.757 MiB/s (took 0.150017 seconds)...........copy64: 39.285 MiB/s (took 0.220559 seconds)
copy64f: 55.978 MiB/s (took 0.154784 seconds)..........copy64f: 38.405 MiB/s (took 0.225609 seconds)
copy64x2: 57.679 MiB/s (took 0.150220 seconds)........copy64x2: 39.285 MiB/s (took 0.220557 seconds)
copy64fx2: 57.878 MiB/s (took 0.149703 seconds).......copy64fx2: 39.327 MiB/s (took 0.220323seconds)
copy64fx2PF: 57.847 MiB/s (took 0.149784 seconds)...copy64fx2PF: 39.290 MiB/s (took 0.220526seconds)
copy64fx4PF: 57.873 MiB/s (took 0.149717 seconds)...copy64fx4PF: 38.521 MiB/s (took 0.224932 seconds)
useMemcpy: 101.914 MiB/s (took 0.085018 seconds)...useMemcpy: 103.002 MiB/s (took 0.084120 seconds)
useExecCopyMem: 101.907 MiB/s (took 0.085024 s)....useExecCopyMem: 103.005 MiB/s (took 0.084118 seconds)
copyToVRAM: 55.979 MiB/s (took 0.154783 seconds)..copyToVRAM: 39.280 MiB/s (took 0.220586 seconds)
WritePixelArray: 103.528 MiB/s (took 0.083693 sec).....WritePixelArray: 103.147 MiB/s (took 0.084002 seconds).

Copy from VRAM to RAM:

copy32: 27.113 MiB/s (took 0.319570 seconds).................copy32: 19.691 MiB/s (took 0.440016 seconds)
copy64: 29.123 MiB/s (took 0.297516 seconds).................copy64: 21.107 MiB/s (took 0.410510 seconds)
copy64f: 31.129 MiB/s (took 0.278344 seconds)................copy64f: 22.470 MiB/s (took 0.385610 seconds)
useMemcpy: 29.946 MiB/s (took 0.289340 seconds)..........useMemcpy: 21.134 MiB/s (took 0.409989 seconds)
useExecCopyMem: 29.956 MiB/s (took 0.289241 sec).......useExecCopyMem: 21.425 MiB/s (took 0.404411 seconds)
copyFromVRAM: 35.935 MiB/s (took 0.241120 seconds)....copyFromVRAM: 24.170 MiB/s (took 0.358477 seconds)
ReadPixelArray: 35.912 MiB/s (took 0.241270 seconds).....ReadPixelArray: 24.518 MiB/s (took 0.353399 seconds).

FillRect:
Size Time (s) Ops/s MPixel/s............................................Size Time (s) Ops/s MPixel/s
(16, 16) 0.035 283029.548 69.099....................................(16, 16) 0.329 30381.285 7.417
(32, 32) 0.059 170404.199 166.410...................................(32, 32) 0.319 31371.663 30.636
((64, 64) 0.157 63628.208 248.548...................................(64, 64) 0.314 31814.610 124.276
(128, 128) 0.427 23416.626 365.885...................................(128, 128) 0.329 30363.205 474.425
((256, 256) 1.325 7545.217 471.576...................................(256, 256) 0.340 29445.714 1840.357
(512, 512) 4.783 2090.561 522.640...................................(512, 512) 1.052 9503.201 2375.800
(1024, 1024)19.337 517.149 517.149...................................(1024, 1024)3.936 2540.941 2540.941

BltBitMap:
Size Time (s) Ops/s MPixel/s .........................................Size Time (s) Ops/s MPixel/s
(16, 16) 0.050 201146.535 49.108 ...................................(16, 16) 0.320 31294.889 7.640
(32, 32) 0.075 133365.341 130.240...................................(32, 32) 0.336 29801.729 29.103
(64, 64) 0.220 45538.792 177.886...................................(64, 64) 0.325 30762.700 120.167
(128, 128) 0.684 14625.977 228.531...................................(128, 128) 0.321 31194.630 487.416
(256, 256) 2.517 3973.370 248.336...................................(256, 256) 0.372 26883.527 1680.220
(512, 512) 9.733 1027.381 256.845..................................(512, 512) 1.075 9302.499 2325.625
(1024, 1024)38.590 259.074 259.074..................................(1024, 1024) 4.026 2484.158

OverlappedBltBitMap:
Size Time (s) Ops/s MPixel/s ..................................................Size Time (s) Ops/s MPixel/s
(16, 16) 0.081 246041.803 60.069...................................(16, 16) 1.515 13204.414 3.224
(32, 32) 0.164 122148.594 119.286...................................(32, 32) 1.616 12377.540 12.087
(64, 64) 0.471 42465.460 165.881...................................(64, 64) 1.637 12220.046 47.735
(128, 128) 1.432 13963.165 218.174...................................(128, 128) 1.655 12081.533 188.774
(256, 256) 5.117 3908.805 244.300...................................(256, 256) 1.718 11643.036 727.690
(512, 512) 19.483 1026.556 256.639...................................(512, 512) 2.877 6951.674 1737.918
(1024, 1024)76.737 260.631 260.631...................................(1024, 1024)9.809 2038.921 2038.921

Composite:
Size Time (s) Ops/s MPixel/s ..................................................Size Time (s) Ops/s MPixel/s
(16, 16) 0.264 37893.716 9.251.....................................(16, 16) 0.406 24654.711 6.019
(32, 32) 0.419 23853.369 23.294...................................(32, 32) 0.409 24471.238 23.898
(64, 64) 0.983 10170.074 39.727...................................(64, 64) 0.395 25296.539 98.815
(128, 128) 3.165 3159.434 49.366...................................(128, 128) 0.401 24908.710 389.199
(256, 256) 12.060 829.196 51.825...................................(256, 256) 0.557 17951.033 1121.940
(512, 512) 52.079 192.016 48.004................................... (512, 512) 1.834 5451.623 1362.906
(1024, 1024) 203.857 49.054 49.054...................................(1024, 1024) 6.995 1429.653 1429.653

CompositeSrcMask:
Size Time (s) Ops/s MPixel/s ..................................................Size Time (s) Ops/s MPixel/s
(16, 16) 0.282 35465.772 8.659...................................(16, 16) 0.437 22890.838 5.589
(32, 32) 0.459 21777.952 21.268...................................(32, 32) 0.432 23167.293 22.624
(64, 64) 1.059 9439.136 36.872...................................(64, 64) 0.446 22402.638 87.510
(128, 128) 3.417 2926.611 45.728...................................(128, 128) 0.433 23096.502 360.883
(256, 256) 12.890 775.768 48.485...................................(256, 256) 0.594 16839.014 1052.438
(512, 512) 55.847 179.061 44.765................................... (512, 512) 1.904 5250.907 1312.727
(1024, 1024)223.099 44.823 44.823...................................(1024, 1024) 7.291 1371.593 1371.593

Random:
Time (s) Ops/s MPixel/s ...............Time (s) Ops/s MPixel/s
376.538 212.462 71.58..................98.150 815.081 274.627

Overall: 992.06............................Overall: 4376.28



Cow3D Warp3D: 35 fps.................Cow3D Nova: 250 fps


shaderjoy 1.19: xxxxxxx...........................shaderjoy 1.19 (project_test.sjp: window, size 800x600): 30 fps

SantasMonsterShootout
:
R9250: 39-49 fps............HD7750: 249-262 fps

Movies. No UVD yet

Small video:

R9250:
VO: [p96_pip] 480x270 => 480x270 Planar YV12

BENCHMARKs: VC: 99.801s VO: 27.742s A: 0.000s Sys: 1.719s = 129.262s
BENCHMARK%: VC: 77.2081% VO: 21.4619% A: 0.0000% Sys: 1.3300% = 100.0000%
CPU do 87%

HD7750:
VO: [comp_yuv2] 480x270 => 480x270 Planar YV12

BENCHMARKs: VC: 100.925s VO: 21.162s A: 0.000s Sys: 1.885s = 123.971s
BENCHMARK%: VC: 81.4101% VO: 17.0698% A: 0.0000% Sys: 1.5202% = 100.0000%
Internal COMP YUV2 - FPS 50
CPU do 83%

HD video:

R9250:
VO: [p96_pip] 1280x720 => 1280x720 Planar YV12

BENCHMARKs: VC: 747.945s VO: 192.421s A: 0.000s Sys: 2.668s = 943.034s
BENCHMARK%: VC: 79.3127% VO: 20.4044% A: 0.0000% Sys: 0.2829% = 100.0000%

HD7750:
VO: [comp_yuv2] 1280x720 => 1280x720 Planar YV12

BENCHMARKs: VC: 726.262s VO: 149.076s A: 0.000s Sys: 3.171s = 878.510s
BENCHMARK%: VC: 82.6698% VO: 16.9692% A: 0.0000% Sys: 0.3610% = 100.0000%
Internal COMP YUV2 - FPS 7

...so, was it worth it? YEAH! . And you know what? Since geennaam has probably removed his mighty Radeon from his SAM440, according to GfxBench2D site, I have the fastest SAM440 on the planet!

PS. Waiting for UVD support for SI

@mr2

Thanks for the report, anyway I wonder how does Basilisk II work on your system.

@mr2

I have been using a Sam440 (M9) many years so was very happy to see Shaderjoy results on this little machine. Have you tried Fractal Nova?

@LiveForIt

Will try Basilisk later. Downloaded ROMs but have set the MacOS8.1 up.


@Capehill

Default window size:

Manderbrot 570 fps,
Julia1 488,
Julia2 483,
Julia3 673,
Julia4 498,
Julia5 799,
Julia6 579,
Julia7 575,
Julia8 516.

good results. The compositing result basically confirms the better workbench experience.

But I have to disappoint you. Someone else uploaded a >25% better result today as well. (Wasn't me.) link

@geennaam

And I have been knocked of the top after about 7 years .
Full list of Sam440ep results.
http://hdrlab.org.nz/benchmark/gfxben ... aOS/Motherboard/Sam440EP/

@mr2

Not bad!

Regarding Shaderjoy, if you want to benchmark, make sure to disable VSYNC because it's enabled by default.

@mr2
Will be also interesting if you can try some of my ports too: H-Craft (press f11 for FPS), SuperTuxKart (there fps enabled by default) , Nights of the Zombies (in window mode FPS in the title of the window), "Worlds" (in window mode FPS in the title of the window) and NeverBall (there i do not remember if FPS is enabled by default or not).


@Capehill
Yeah, I was about to say the same: 30 fps mean he has VSYNC enabled. But then it also means, that it for sure not more than 60 (as when rendering is less than 60 in shaderjoy it then switches to 30 if can reach it too).

@geennaam
Quote:
Drat!

@Capehill
Quote:
I knew that...and forgot to turn it off
Now, the result is 38/39fps

@kas1e

Quote:

That was the plan but I'm in the middle of a painfull process of configuring Basilisk and installing Mac OS8.1. I have lost few hours just to discover that ROM and MacOS are OK but I can't see
an initial sreen because of the wrong display setting.
CPU is 100% so, I will be able to do more tests maybe in a week or two

@mr2

I never managed install anything else then MacOS7.6, disable FPU emulation, it’s not good. Is there point to installing Mac8/9, is not most of that stuff PPC?

So what's the best gfx card for 440 / 460 ?

@geennaam
Quote:

It was me.
According your advice I tried other PCI-PCIe bridge - P17C9X 111SLBFDE.
The card is Powercolor Devil R9 270X ( borrowed from my AmigaOne X1000) and my Sam440ep-flex is 733MHz version.
This card is dual-slot and 180 TDP, a little bit overkill for little Sam .

And I have no succsess with PCI-PCIe bridge PEX8111 (and have no 8112). With 8111 works only U-boot, in serial debug is
RadeonHD.chip (0): Connector[4] DVI-D 2:
RadeonHD.chip (0): Type: RHD_CONNECTOR_DVI_SINGLE
RadeonHD.chip (0): DDC bus: RHD I2C line 4
RadeonHD.chip (4): rhdSIHPDCheck returned: 100 mask: 100
RadeonHD.chip (0): HPD: RHD_HPD_1 (monitor connected)

but black screen. I have to experiment with cables and connectors more.

@NinjaCyborg

Just look at the post #6.

@LiveForIt

So I can say that Basilisk works very slow. The only working displays are B&W (Window and Window composition) with the internal convertion. I can turn on the colors from MacOS prefs, after loading the OS. All other displays don't work for me. Sometimes MacOS loads (I can hear a notification sound) but there is no display, sometimes it crash. Overall it is too slow (CPU 99/100%) to use on my SAM. Unfortunately after all hard crashes I have to turn off the power for several minutes because if I do it faster (e.g. 15/20 s) SAM does not boot.


@kas1e

Night Of The Zombies runs at 7 fps-is it possible to lower the resolution?
foobilard++ is very playable and looks very good, prevoius version was unplayable with R9250

more to come...

@sailor

...so, you were hiding

@sailor

From what i've read in other threads, the RadeonHD driver tries to improve memcopy speed with changing settings in PEX811x bridges. So maybe that's the source of the issue. It was reason for me to try another bridge.

PEX8112 cards are hard to find anyways. People who tried to order one ended up with a PEX8111 anyways.

@mr2
Quote:

It's hardcoded sadly, but 7 FPS on sam440 is not that bad for this game! It only 30 on x5000, so pretty good IMHO.

Comparing the Same Card Model across different Machines

http://hdrlab.org.nz/benchmark/gfxben ... AmigaOS/GraphicsCard/568/

http://hdrlab.org.nz/benchmark/gfxben ... AmigaOS/GraphicsCard/321/

http://hdrlab.org.nz/benchmark/gfxben ... AmigaOS/GraphicsCard/312/

@mr2
Quote:
I am new here. I am long-time member on Hyperion and Morph.zone forums.
And I have my Sam three weeks only.

@geennaam
Quote:
I think, you are right.

I check the debug output, and with use of PEX8111 bridge, there is allways error with AtomBios

then Radeon recognized correctly conectors and connected monitor, and black screen.


With PI7C9X bridge Atombios works normally.

@mr2

Thanks for the report.
100% usage is normal as 68k is emulated, but setting low CPU priority, should make Basilisk behave friendly. Your problems with no graphics are worrying. I had no report of problem yet. WPA mode is expected to be bad, just because it requires MacOS modes compatible with AmigaOS modes, not all are.

@NinjaCyborg

Quote:

One that's confirmed to be working with the configuration you have. geennaam tested an MSI R9 270X Gaming and Spectre660 said the VisionTek Radeon 7750 SFF worked in his Sam440 so if you can track down one of those and use the same adapters it should work for you too.