Overclocked Radeon Card Breaks 1 GHz

dacaldar writes "According to Yahoo Finance, noted Finnish over-clockers Sampsa Kurri and Ville Suvanto have made world history by over-clocking a graphics processor to engine clock levels above 1 GHz. The record was set on the recently-announced Radeon® X1800 XT graphics processor from ATI Technologies Inc."

Re:Benchmarks?

[Mr. Sketch]

Graphic showing 3DMark score of 12419:
http://www.muropaketti.com/3dmark/r520/12419.png [muropaketti.com]

Pictures of their setup/methods:
http://www.muropaketti.com/3dmark/r520/ghz/ [muropaketti.com]

This record already broken

[Ezku]

Sampsa and Ville already broke their own record by overclocking the same setup to over 1GHz for both the GPU and memory. See pictures over at Muropaketti [muropaketti.com].

Re:GPU to excel CPU

[LLuthor]

http://www.gpgpu.org/ [gpgpu.org]

Re:GPU to excel CPU

[OzPeter]

You mean like these people are doing?

Generic GPU programming [gpgpu.org]

Re:GPU to excel CPU

[Jerry Coffin]

The performance of GPU's seem to grow faster than those of CPU's. I remember someone had proposed to use GPU's to proces generic data. It would be 12 times faster than a CPU.

Go here [gpgpu.org] for several examples of this -- far from simply having been proposed, it's been done a fair number of times.

The thing to keep in mind with this is that while the GPU has a lot of bandwidth and throughput, most of that is due to a high degree of parallelism. Obviously 1 GHz hasn't been a major milestone for CPUs for quite a while, but CPUs are only recently starting to do multi-core processing, while GPUs have been doing fairly seriously parallel processing for quite a while.

Along with that, the GPU has a major advantage for some tasks in having hardware support for some relatively complex operations that require a fair amount of programming on the CPU (e.g. multiplying, inverting, etc., small vectors, typically has a single instruction to find Euclidean distance between two 3D points, etc.)

That means the GPU can be quite a bit faster for some things, but it's a long ways from a panacea -- you can get spectacular results applying a single mathematical transformation to a large matrix, but if you have a process that's mostly serial in nature, it'll probably be substantially slower than on the CPU.

Along with that, development for the GPU is generally somewhat difficult compared to development on the CPU. Writing the code itself isn't too bad, as there are decent IDEs (e.g ATI's RenderMonkey) but you're working in a strange (though somewhat C-like) language. Much worse is essentially a complete lack of debugging support. Along with that, you have to take the target GPU into account in the code (to some extent). I just got a call in the middle of a meeting this morning from one of my co-workers, pointing out that some of my code works perfectly on my own machine, but not at all on any his. I haven't had a chance to figure out what's wrong yet, but I'm betting it stems from the difference in graphics controllers (my machine has an nVidia board but his has Intel "Extreme" (ly slow) graphics).

--
The universe is a figment of its own imagination.

Re:GPU vs. CPU Speed

[freidog]

Since DirectX 8 (I think), the color values have been floating point numbers, this is to avoid loosing a lot of possible values through all blending with multi-texturing and effects (fog, lighting ect) which are of course much slower than very simple integer calculations. Even on the Athlon64's FP add and muls are 4 cycles, you'd have to make the top end A64 about 700mhz if you make them single cycle execution. (multi-cycle instructions aren't as bad a thing on the CPU as there are plenty of other things to do while you wait, not so in GPUs).

GPUs have also tended to focus on parallel execution - at least over the last few years - increasing the number of pixels done at the same time, to compensate for not being able to hit multi-ghz speeds, so yes they have many more transistors than typical CPUs (the 7800GTX might break 300 million, well over 250 million) - and of course heat is an issue if you push the voltage and / or clock speeds to far. The last few generations of GPUs have been up around 65-80W real world draw, more than most CPUs out there. And of course GPUs have very little room for cooling in those expansion slots.

Because it's worth zilch without...

[GrAfFiT]

...the pictures of the rig : here they are [muropaketti.com], 3DMark05 included [muropaketti.com].

O RLY?

[irc.goatse.cx troll]

http://www.bfgtech.com/7800GTX_256_WC.html [bfgtech.com]

BFG GeForce(TM) 7800 GTX OC(TM) with Water Block. Factory overclocked to 490MHz / 1300MHz (vs. 400MHz / 1000MHz standard), this built-to-order card will feature a water block instead of a GPU fan for those wanting to purchase or who may already have an existing liquid-cooled PC system. BFG will hand-build your card using Arctic Silver 5 Premium Thermal Compound. Easily hooked up to any existing 1/4" tubing system or to 3/8" tubes with the included adapters, this card runs cool and silent. BFG Tech is proud to offer their true lifetime warranty on this graphics card. (Card with water block requires internal or external water cooled system, sold separately.)

Re:One wonders...

[Eugene]

that's 1800XL, (comparable to 7800GT). 1800XT is probably still a month away from release.

 

Re:GPU to excel CPU

[Mostly a lurker]

A good question! This excerpt from a recent article in Extreme Tech [extremetech.com] seems relevant:

The third future project at ATI is dramatically improved support for the GPGPU scene. These are researches, mostly academic, that are tapping into the massive parallel computing power of graphics processors for general computing tasks, like fluid dynamics calculations, protein folding, or audio and signal processing. ATI's new GPU architecture should be better at GPGPU tasks than any that has come before, as it provides more registers per pipeline than either ATI's old architecture or Nvidia's new one. This is a sore spot for GPGPU developers but not really a limitation for game makers. The improved performance of dynamic branching in the new architecture should be a huge win for GPGPU applications as well. Developers working to enable general purpose non-graphics applications on GPUs have lamented the lack of more direct access to the hardware, but ATI plans to remedy that by publishing a detailed spec and even a thin "close to the metal" abstraction layer for these coders

Re:Benchmarks?

[diablomonic]

have a look at xtremesystems.com/forums, this is where they talked about it first (I was reading it there a couple of days ago). at that stage they had graphics core at 1.0 something ghz and memory at 2.0 something ghz, but it was only stable in 2d mode. the highest they could get in 3dmark at that stage was around 12400 and yes, that was with the overclock backed off a bit to 800 and something

in other words... still impressive (no other chip has been able to overclock to 1ghz, even in 2d mode) but not quite what you were hoping for

Re:Huzzah!

[Anonymous Coward]

note that before you go celebrating too much, the 1ghz speed was only stable in 2D MODE: the 3dmark score is for when they had backed off the overclock to about 880, see here

http://www.xtremesystems.org/forums/showthread.php ?p=1104977#post1104977 [xtremesystems.org] sampsa is asked Were you able to run any benchmarks at that speed or was that a Windows stable shot???? Anyway that is still hella fast with no artifacts. sampsa's response Just a Windows shot in 2D.

so still impressive, but not what they describe