Dave Baumann Saves the Radeon HD 4850

ATI had this habit of finding good reviewers and bringing them on staff. Our first Graphics Editor, Matthew Witheiler, went to work for ATI after graduating from Duke. He was with AnandTech for a good three years before ATI snagged him, he ended up being ATI’s youngest Product Manager (congrats on the engagement Matthew). One other prominent reviewer ATI grabbed ahold of was Dave Baumann of Beyond3D fame and brought him on to do technical marketing.

One of Baumann’s strongpoints was the ability to analyze the competitive landscape given that’s what he did for Beyond3D before ATI. One of Dave’s first major tasks at ATI was to compare R600 to G80 internally, which wasn’t exactly the best job in November of 2006. Obviously, G80 had a significant impact on RV770. While the architecture was set in stone, clock speeds, board layout and memory sizes were all variable until early 2008.

Initially, RV770 was targeted at 1.5x the performance of R600, which looking back would not have been enough. During the next 1.5 years that 1.5x turned into 2x R600 and finally settled at 2.5x the speed of R600, at a price in the $200 - $300 range.

Dave became a product manager on RV770 by February 2008, which was a big deal given that he hadn’t been with ATI that long and this was a very important product. RV670 saw ATI return to competition in the year prior, but RV770 needed to put ATI back on top.

When Dave took the 770 under his wing a lot of the product had already been mapped out, the chip was back from the fabs and at this point ATI’s engineering team wasn’t ready or eager to make any changes. The RV770 XT sat well with Mr. Baumann (the XT was the internal name of the Radeon HD 4870), in his words “the specifications were perfect”. There was a late change to the 4870 that gave it its second PCIe power connector, but that’s it. Arguably the more important version, the RV770 Pro that would become the Radeon HD 4850, concerned him - it was a bit under spec’d.

Here’s a quick put-yourself-in-ATI’s-shoes test. Your engineering team has spent the past three years on a product that may fail miserably because it’s a radical departure from how you’ve designed GPUs in the past. Your last major GPU architecture launch failed miserably (R600), and the last refresh (RV670) did ok but still didn’t really snag real mindshare from NVIDIA. You’ve just finished this radical new design, and this young new PM with an accent comes in three months before you’re supposed to enter production and tells you that you need to make changes. It was a ballsy move by Baumann, but he wasn’t interested in saving face, he was trying to help his team win. The engineers could’ve just as easily cast him aside, but they listened and they worked, oh did they work. The final stretch is rarely the quickest or the easiest, and this is very true about RV770.

The Radeon HD 4850 was originally a 256MB card with a 500MHz core clock and 900MHz memory clock. Dave insisted that the card needed 512MB of GDDR3 and 625MHz core / 993MHz memory clock, it’s not just that he insisted, but that he convinced the engineers to make such a late change. Dave took the engineers through his reasoning of why and where ATI needed to be in the competitive landscape, by the end of the discussion he didn’t need to persuade them, the board and ASIC teams were championing the changes.

Had it not been for these modifications, the 4850 would not have put as much pressure on NVIDIA’s GeForce 9800 GTX and its pricing wouldn’t have needed to fall so quickly.

Thanks Dave.

Just One Small Problem: We Need a New Memory Technology The Last Hiccup and Recon from Taiwan
Comments Locked

116 Comments

View All Comments

  • DerekWilson - Thursday, December 4, 2008 - link

    I'm glad you guys were able to stick to the plan an launch at the amazing prices you hit. It really shook up the industry and helped bring higher performance to lower price points. Now we just need the same thing to happen with integrated graphics.

    But seriously ... about those future architectures ... maybe you guys want to sit down and have another nice long chat? ;-)
  • tygrus - Thursday, December 4, 2008 - link

    Choice A : Spend another 1B on a larger chip R&D, fab problems, low yields, high unit cost, potential delays, large dual-slot with space&cooling problems. Loose mid-range revenue potential, more R&D/time etc. to make half-size chip for mid-range.

    Choice B : Make the design to be price/power/performance efficient for very profitable mass-market. If you users want almost double the performance, buy 2 identical cards which saves on not producing a low volume BIG chip/card. (Though now we have 2 chips for X2 which still simplifies and reasonable efficient compared to a card design cramming a single BIG chip starved of bandwidth).
  • CloudFire - Thursday, December 4, 2008 - link

    hands down one of the best, if not the best article i've ever read at anand. GREAT JOB! it shows how believing and getting back up when you're being beat down can get you where you want to be!

    for those people who took up the challenge to be different and change, imo are the true innovators who changes the world in the face of extreme adversity.
  • Emperor88 - Thursday, December 4, 2008 - link

    Great artilce. There are precious few of these.

    Thanks a lot :)
  • cowofdoom - Thursday, December 4, 2008 - link

    Long time reader, first time poster. Great article. I really think it says it all. Any time you get a chance to write articles like this I would love to see them. Great job.
  • Emperor88 - Thursday, December 4, 2008 - link

    haha first time (wait second now) poster here as well. Best article I've read here.
  • Barack Obama - Thursday, December 4, 2008 - link

    Keep it comin'!
  • Boundless - Wednesday, December 3, 2008 - link

    Very nice article once again...helps reinforce my decision of keeping this place readily bookmarked since I first visited back in 2001.
  • Regs - Wednesday, December 3, 2008 - link

    I want to know what AMD was thinking for those quite 3 years when they blew Prescott out of the water. Did they see Conroe coming? I think it's time for Anand to get on some black paint and go commando over there at AMD HQ.
  • Cloudie - Wednesday, December 3, 2008 - link

    One of the best articles I've ever read. Kudos to Anand, and thanks to AMD (:

Log in

Don't have an account? Sign up now