So I got my Gigabyte GeForce GTX460OC 1 GB card

And this is what it looks like:

It’s the first product from Gigabyte that I ever owned, by the way.

Well, first impression of the card is quite good.
It really is incredibly silent compared to the Radeon 5770. It’s much like the 8800GTS I used to have… the fans have a barely audible low hum. I mainly hear the large fan in my PSU spinning up when the card is under load. With the Radeon 5770, the fan had a nasty whine to it, which could be heard over all other sound.
What’s worse: the 5770 apparently boots into full 3D clocks. Even though there is some fan control, if you stay in the POST screen too long (that is, more than say 30 seconds), the fan will spin up like crazy (probably a sign that the GPU is also running at very high temperature). And if you boot into an OS without the proper drivers installed, it will stay like this all the time. Only when the driver is loaded to control the fans and GPU speed, the noise finally dies. Not nice when you boot into a linux live CD with open source drivers, which don’t have this fan control…

It also fixes some gripes regarding drivers… Firstly I don’t have the problem anymore that when I do a remote desktop session to my machine, that the Catalyst Control Panel dies on me. Typical unprofessional bug in ATi’s drivers, has been there as long as I’ve had the card (since November last year), and was never fixed.
Secondly, OpenGL works, and works well. With the 10.7 Catalysts, the Unigine Heaven benchmark would just hang your entire PC with a black screen (even though the release notes specifically say that this has been fixed). And with older drivers, although Heaven worked, the OpenGL performance was considerably worse than DX11, as I mentioned.

Initially I thought that OpenCL was broken on the GTX460, because every OpenCL application I tried, just crashed. Then I realized what the problem was: Since OpenCL isn’t installed with the Catalyst drivers, it is not UNINSTALLED either. The OpenCL runtime is actually part of the Stream SDK, and ATi sets some environment variables that point to the binaries. So it was still pointing to these OpenCL binaries rather than the nVidia ones (which were properly installed in system32 and syswow64). After manually cleaning out my path variable, things started to work… and work well.
I guess I don’t have to worry about OpenCL anymore, it just works and comes with the standard drivers. And they bother to test it too..
The Catalyst 10.7 release notes stated that OpenCL was not tested for conformance in these drivers, and they recommended going back to 10.5 if you needed OpenCL… great… while at the same time they release a marketing campaign like this:

Our promise to nurture open industry standards
PC game developers have a right to the best possible gaming platform. AMD’s Gaming Evolved program confirms our promise to deliver the most innovative technologies, tools, and industry support to maintain the PC platform as world’s premier gaming environment. We will participate in the development and cultivation of OpenCL and OpenGL industry standards, and we will move quickly to move our innovations into the industry standards whenever feasible.

Okay… so you promise to ‘nurture’ open industry standards like OpenGL and OpenCL… but at the same time you release a driver where OpenGL literally hangs your PC so badly that powering off is the only option… and you don’t even bother to test the OpenCL portion in your drivers, and recommend using older drivers instead?
Right… I think I’m going to be very happy to have gotten rid of my Radeon 5770. Seeing as I like to use open industry standards like OpenGL and OpenCL.

Oh, and I played a small game of Mirror’s Edge. Running with PhysX for the first time since my 8800GTS died. Still pretty cool to have the extra effects.
I was thinking… this is the second time where I actually started to appreciate a brand more because I was using their competitor’s stuff.
The first time was with my Athlon. Okay, it had a VIA chipset, so you can’t blame AMD for everything (except perhaps for too low supply of their own chipsets, forcing many people to buy inferior third-party chipsets), but it was a total disaster to get that thing stable… and even when it finally was stable, it ran very hot and noisy. My brother bought an Athlon as well at that time, and he eventually gave up, chucked the lot and bought a Pentium 4 and Intel chipset. Although I stuck with the AMD system myself, I could understand his move. You really learn to appreciate the qualities of Intel’s chipset, power management, cooling solution etc when you’ve been using an AMD system for a while.

Same thing now… Although I’ve had various ATi cards before, and never really had a problem with them at the time (Radeon 8500, Radeon 9600, X1900), I guess the situation is slightly different today.
I had gotten used to my 8800GTS, with good OpenCL support, Cuda, PhysX, and everything. Nice and silent too.
So the ATi card was a bit of a letdown. I had to give up on a lot of stuff, and also ran into various driver bugs… Another thing… if I made a bug in a D3D shader, causing the app to send geometry without setting a valid shader, the driver/GPU would often crash. Which would take a while to recover, very annoying. I haven’t tested that yet, but I never had that problem with the 8800GTS at the time. I hope the GTX460 is just as robust.

I did find a slight glitch though, perhaps driver-related… The DirectX 11 subdivision sample does not work. One of the hull shaders compiles (the SubDToBezierHS4444 function), but when you try to create a shader object from the compiled blob, it returns an out-of-memory error. Oh well… this is just the first set of drivers ever, for the GTX460. I’ve learnt to accept that nobody seems to be able to get it right the first time (the DX10 drivers for the 8800GTS weren’t without glitches either, initially). It’s just that I drew the line with the Radeon 5770, where certain things STILL did not work after more than half a year, and 6-7 updates of the driver… and in some cases, things that actually worked earlier, were broken again (like the Unigine Heaven benchmark in OpenGL). And not even bothering to test the OpenCL portion of the driver? I find that unacceptable, for an official release driver anyway. That’s going to be very difficult for when OpenCL applications are released as well… The publishers will have to explain to the end-users that they cannot just upgrade their drivers to the latest (while in most other cases, that is exactly what you would advise). They have to make sure that they use drivers that have been tested for compliance with OpenCL. Which apparently AMD hasn’t done since the 10.5 release.

This entry was posted in Uncategorized. Bookmark the permalink.

One Response to So I got my Gigabyte GeForce GTX460OC 1 GB card

  1. Pingback: nVidia stability issues on GeForce 400/500 series | Scali's OpenBlog™

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s