Interview with John Bruno of ATI - Fish and chips? Oh yes.

[Steve]

We talked about things we shouldn't have talked about, but plenty of things I can relate back to you so you can get an insight into John, what his team does and where ATI are headed with core logic. They have a grand plan with the lofty goal of dethroning NVIDIA from the top of the enthusiast mainboard space. John wants you to be buying ATI-based boards for your high-end and overclocking systems, instead of nForce4, and he outlined why he think that'll happen.

Go fish

[S1fter]

very interesting, it will always be a cat and mouse game with regards to performance by both ATI and nVidia.

Any info with regards to new up and comming GPU's (Come on tell us)

[chrestomanci]

I am amazed at John's claim of getting 3D mark working on fresh silicon from the fab in only nine minutes.

I used to work for another fabless semiconductor company, and I would say that 9 weeks would be a more normal figure. Even if the silicon is fine, all sorts of other stuff can hold things up.

There are any number of simple issues, like voltages, memory timing, bugs in the bios, and any number of electrical design problems that can hold stuff up for a few days or a week. On one project that I worked on, I recall the dev board the chip was plugged into was constantly having resistors changed to correct various issues, and logic probes attached to investigate them.

The only way I can imagine getting something working in nine minutes, would be if this was a re-spin of something that was nearly working to begin with, and everything else was already debugged and working.

[Dougal]

Basicly I think the 9 minute claim is after all there initial testing.

They get the Chip back, solder it onto a almost complete board, cool it.

Set the system up and install.

9 mins is possible I think.

External setup, ghost the drive.

[Steve]

Originally Posted by Dougal

External setup, ghost the drive.

It'll already be loaded with the OS from previous testing, I'd wager.

[JBruno]

I see this is someone with some real experience at doing new ASIC bring ups. But let me confirm that this was not a respin. It was a new process node and some significant logic changes. To make something like this happen (bring up a new chip in less than 10 minutes) takes a lot of planning and a lot of verification.

Originally Posted by chrestomanci

I am amazed at John's claim of getting 3D mark working on fresh silicon from the fab in only nine minutes.

I used to work for another fabless semiconductor company, and I would say that 9 weeks would be a more normal figure. Even if the silicon is fine, all sorts of other stuff can hold things up.

There are any number of simple issues, like voltages, memory timing, bugs in the bios, and any number of electrical design problems that can hold stuff up for a few days or a week. On one project that I worked on, I recall the dev board the chip was plugged into was constantly having resistors changed to correct various issues, and logic probes attached to investigate them.

The only way I can imagine getting something working in nine minutes, would be if this was a re-spin of something that was nearly working to begin with, and everything else was already debugged and working.