@Karlos
I'm no Windows fan, but professional companies such as Autodesk or Epic do find ways to make stuff run faster on windows than any competing solution on any other system.

But if you tell me GPGPU is efficient on linux I'm interested to hear about it, can you make some examples? (genuinely interested)

_________________
SamFlex Complete 800Mhz System + AmigaOS 4.1 Update 4
Amiga 2000 DKB 2MB ChipRam GVP G-Force040 Picasso 2 OS3.9 BB2
AmigaCD 32

@DAX

Not sure about GPGPU on all hardware architectures, but CUDA on linux kicks #### plain and simple. I've used it on both linux and windows and although the tools available for windows are perhaps nicer (better integration with VS etc), raw performance is hard to beat on linux. I have no idea why it should be the case other than the fact that perhaps linux is just more efficient at dealing with high load tasks in general. I've also found it to be more stable for long running compute tasks and better at distributing load between multiple GPUs.

In fact, the Fastra II (http://fastra2.ua.ac.be/), which AFAIK is the fastest "desktop" GPGPU computer around is only made possible via Linux. It has 13 G200 class GPUs which required a custom motherboard bios just to allow them to be mapped in and they've had to modify the kernel to allow all the cards to be accessed. You won't get away with that sort of thing on Windows :)

_________________
Doing stupid things for fun...

@Karlos
Cool

_________________
SamFlex Complete 800Mhz System + AmigaOS 4.1 Update 4
Amiga 2000 DKB 2MB ChipRam GVP G-Force040 Picasso 2 OS3.9 BB2
AmigaCD 32

@hazydave

Quote:

MacOS is a religion.

Well, not a very good one — their messiah has a health crisis every other year.

_________________
"Unix is supposed to fix that." -- Jay Miner

Quote:

TrevorDick wrote: @nimrod7 Quote: but i was referring to NEMO motherboardb Oh OK. NE(w) MO(therboard) = NEMO TrevorD

I didn't know that, or had worked out the anagram of A-one...

_________________
Rage quited 29th May 2011

@bison

Quote:

Well, not a very good one — their messiah has a health crisis every other year.

Not a problem. I know one relatively successful religion whose messiah was brutaly executed some years ago.

Quote:

Not a problem. I know one relatively successful religion whose messiah was brutaly executed some years ago.

Someday Steve will be laid to rest (in an aluminum casket), and that will be the end of the cult of Mac. I don't think Apple will successfully transition to Steve++, whoever that may turn out to be.

Last edited by bison on 21-Apr-2011 at 10:30 PM.

_________________
"Unix is supposed to fix that." -- Jay Miner

@bison

Quote:

and that will be the end of the cult of Mac.

Cult is based on Faith, Prophet is only Messenger. Faith will find new Messenger.

@bison

Quote:

I don't think Apple will successfully transition to Steve++, whoever that may turn out to be.

What if Steve++ turned out to be an iClone ?

_________________
Doing stupid things for fun...

@pavlor
not that easy it would seem though, when Jobs left the last time, Apple fell into oblivion (until he returned, "The Second Coming of Jobs" )

_________________
SamFlex Complete 800Mhz System + AmigaOS 4.1 Update 4
Amiga 2000 DKB 2MB ChipRam GVP G-Force040 Picasso 2 OS3.9 BB2
AmigaCD 32

@itix

Quote:

itix wrote: @Arko Quote: It was a in a code they wanted sell to Amiga Technologies, and Andy Finkel was send to Phase5 to do some code review. Yes and? Where is the problem?

Don't know if there is a problem, you mentioned the code might be send to Phase5 by Petro and I told you the reason why that code was not send by Petro.

Quote:

But I wouldnt mind if those old people tried to actually make some sense.

Dave was always a hardware guy, building cool hardware is what he likes to do, that's the reason why he likes the Natami. His comments about AOS4 where not very positive too. In a lot of comments he talks about economic decisions this might be the reason why likes the C64 from CUSA. According to his behavior about MorphOS, maybe there where other reasons, he does not give technical reasons, he does not talk about GUI design he just said he doesn't care.

I'm more surprised why he doesn't try to participate in the Minimig project. Today's hardware design technologies with FPGA makes a lot of thing possible for hobbyists, things that where impossible for Haynie as professional hardware developer 20 years ago.

A lot of people that worked on or with the Amiga, left the scene after their projects failed, they don't want to work with the Amiga again because they went to a lot of disappointments after all this fails.

_________________
AmigaONE. Haha. Just because you can put label on it does not make it Amiga.

I borrowed this comments from here (#27 & #28):
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=38873&forum=2&start=20&order=0

@Karlos

Quote:

Karlos wrote: @DAX Not sure about GPGPU on all hardware architectures, but CUDA on linux kicks #### plain and simple. I've used it on both linux and windows and although the tools available for windows are perhaps nicer (better integration with VS etc), raw performance is hard to beat on linux. I have no idea why it should be the case other than the fact that perhaps linux is just more efficient at dealing with high load tasks in general. I've also found it to be more stable for long running compute tasks and better at distributing load between multiple GPUs. In fact, the Fastra II (http://fastra2.ua.ac.be/), which AFAIK is the fastest "desktop" GPGPU computer around is only made possible via Linux. It has 13 G200 class GPUs which required a custom motherboard bios just to allow them to be mapped in and they've had to modify the kernel to allow all the cards to be accessed. You won't get away with that sort of thing on Windows :)

http://research.microsoft.com/en-us/um/redmond/events/escience2008/matsuoka-escience2008.pdf

1 rack server has 4 gpu cards(8 GPU chips) with Windows 2008 Server.

http://www.excaliberpc.com/593085/asus-p7p55-ws-supercomputer-intel.html
5 PCI-Ex X16 slots ( 10 GPU chips) with Windows 7 support,

Recent Catayst windows 7 drivers supports >32bit address space i.e. issues with DIY ViDock.

Last edited by Hammer on 21-Apr-2011 at 11:27 PM.
Last edited by Hammer on 21-Apr-2011 at 11:05 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Using just 4 graphics cards it not a problem, even for CUDA on Windows. The Fastra-II, however, has 7 cards. ASUS had to provide them with a modified BIOS in order for the hardware to even start up, let alone getting all 7 cards mapped and working in the OS. They were able to do this because linux is open enough for them to make whatever modifications they needed to low level kernel code. Good luck achieving that with closed a source OS.

Either way, I've yet to see any evidence at all that running Windows for this purpose offers any tangible benefits at all, or is even capable of reaching performance parity with linux for the same job.

It's all horses for courses. I'll happily fire up Windows for Gaming but for CUDA, it's of limited use to me. The exact same GPU kernel (same object code produced by NVCC on my machine) takes longer to run and slows down the machine on 64-bit windows than it does on my 64-bit linux install... *shrug*

_________________
Doing stupid things for fun...

@Karlos

Which 64bit Windows? there are differences with 64bit Windows.

Resource allocation issues is the same with DIY ViDock. Some BIOSs are not fully ACPI compiliant e.g ASUS G1S/G1Sn/N80VN CUDA laptops.

There a PCI address defrag boot tool that works with Windows.

The posted Asus i7 motherboard has 5 PEG slots and it's certified for Windows 7.

For each GPU chip, Windows 7 needs 256mb address window to be located beyond 32bit address space i.e drivers must support PAE 36bit.

6 GPU chip(3 cards) setup can easy exceed 1GB system address space (/3gb switch Windows mode).

PS; i only run with AMD Stream GPUs.

Last edited by Hammer on 22-Apr-2011 at 12:12 AM.
Last edited by Hammer on 22-Apr-2011 at 12:05 AM.
Last edited by Hammer on 21-Apr-2011 at 11:49 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Vista 64-bit, in this case, but tested with Windows 7 too, no significant improvement for me.

Quote:

PS; i only run with AMD Stream GPUs.

Boooo!!! Seriously, though, what's your poison on that one OpenCL, DirectCompute, Stream?

_________________
Doing stupid things for fun...

@Karlos

Quote:

Karlos wrote: @DAX Not sure about GPGPU on all hardware architectures, but CUDA on linux kicks #### plain and simple. I've used it on both linux and windows and although the tools available for windows are perhaps nicer (better integration with VS etc), raw performance is hard to beat on linux. I have no idea why it should be the case other than the fact that perhaps linux is just more efficient at dealing with high load tasks in general. I've also found it to be more stable for long running compute tasks and better at distributing load between multiple GPUs. In fact, the Fastra II (http://fastra2.ua.ac.be/), which AFAIK is the fastest "desktop" GPGPU computer around is only made possible via Linux. It has 13 G200 class GPUs which required a custom motherboard bios just to allow them to be mapped in and they've had to modify the kernel to allow all the cards to be accessed. You won't get away with that sort of thing on Windows :)

Relative to RV770, the double float with G200 GPU is pretty slow.

Current multi-GPU PC setups are based on NV Fermi or AMD Cayman.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Karlos

AMD Stream for math array benchmarks and Cyberlink PowerDirector 9. Compute Shader for graphics.

It would be interesting to see console programming on modern GPUs i.e assuming Wii 2 has a cheap fat R700 type GPU.

Anyway, AMD has stated DX gets in the way on pure performance.

Last edited by Hammer on 22-Apr-2011 at 12:40 AM.
Last edited by Hammer on 22-Apr-2011 at 12:37 AM.
Last edited by Hammer on 22-Apr-2011 at 12:33 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

The G200 was the first nVidia GPU to add support for double precision and there aren't as many units for it as there are single precision. If you are going to use it extensively, you opt for Fermi. Finally, the only valid comparisons (as in ones that really matter) between ATI and nVidia for computational power are OpenCL or DirectCompute. There aren't any other serious compute architectures that can run on both.

_________________
Doing stupid things for fun...

@Hammer

PS: For a fun traditional GPGPU (in GLSL) demonstration, you might enjoy the following site:

http://fractal.io/

You need a WebGL capable browser though. Chrome seems to work nicely.

_________________
Doing stupid things for fun...

@Karlos

For 64bit FP, RV770 combines two 32bit stream units. Similar trick to AVX 256bit support with two 128bit FMA units in AMD bulldozer.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)