The AMD Execution Thread [2007 - 2017]

[swaaye]

I seem to recall the VIA KT266a - 333 - 400 chipsets quite popular actually. The only problem was actually some versions of the chip itself.

http://www.geek.com/procspec/amd/thoroughbred.htm

"There's two flavors of the Thoroughbred core, the first hot-running and not-very-scalable Thoroughbred A and the cooler and higher-clocking Thoroughbred B. The "B" core is slightly larger due to a new layout designed to improve scalability"


The chipset problems only occurred with K8 release when VIA chipsets did not have a lockable AGP PCI bus I seem to recall.

I was referring to the Slot A Athlon. AMD's 750 chipset was all you really had at launch. VIA rather quickly released the KX133, though. But the number of decent boards was very limited. ASUS would only release theirs in a basically un-named brown box. It was very slow going for a while there.

VIA's early Athlon chipsets weren't all that great. By early, I mean KX133, KT133, and KT133A. The 686B southbridge was bad news. PCI design was awful. There was occasional IDE data corruption and most of a time a SBLive! would not work correctly. George Breese's PCI Latency patch saved the day there tho, for the most part. The bus had like a pathetic maximum bandwidth of 30 MB/s without his patch. I don't remember how bad their AGP design was, but I bet it had tons of issues too considering how god awful it was on MVP3.

Still, Athlon and Athlon XP turned around AMD in a big way. The K6 days were certainly worse.

 

[overclocked_enthusiasm]

From what I can recall, ATI had a negative cashflow of about $25 million per quarter in Q3 2006. With $500 million in cash they had no real risk of BK. Their share price was about $15 per share before the merger and with the R600 screwup they would be in the $10 per share range right now imho. Orton pulled a rabbit out of his hat getting north of $20 per share to be sure.

http://ir.ati.com/phoenix.zhtml?c=105421&p=irol-newsArticle&ID=877463&highlight=
Liquidity and Financial Resources

On a sequential basis, inventory of $366 million in the third quarter declined 14% from the second quarter of fiscal 2006 reflecting a continued focus on inventory management. Inventory was up from $348 million at August 31, 2005. Days of inventory at quarter end were approximately 79 days based on the previous quarter's sales.

Accounts receivable at quarter end were $436 million as compared with $386 million at August 31, 2005 and reflect higher sales and the timing of payments. Accounts payable and accrued liabilities of $641 million were down from $665 million at August 31, 2005. Both accounts receivable, and accounts payable and accrued liabilities are in line with current revenue levels.

Cash deficit from operations was $26.6 million in the quarter as compared with cash generated of $15.2 million in the third quarter of fiscal 2005 and a cash deficit of $8.6 million in the second quarter of fiscal 2006. The decrease in cash flow from operations from 2005 to 2006 was primarily due to changes in working capital accounts and the timing of payments.

Cash position (cash, cash equivalents and short-term investments) at quarter end was $518 million, down from $587 million at August 31, 2005 and $607 million at February 28, 2006. The sequential decrease reflects the reduction in cash from operations, as well as $24.7 million in share repurchases and $37.2 million on the acquisitions of XGI and Bitboys. At May 31, 2006, we had working capital of $702 million as compared to $660 million at August 31, 2005.

 

[jimmyjames123]

I think perhaps Orton oversold R600 in regards to launch timing and performance. While this may not be any different than the examples you listed above it does differ in that the intellectual property of ATI was the driving force behind the $5.4 billion merger. Given that this intellectual property may not be as valuable as advertised AMD may have felt they were sold a bogus bill of goods and the heat on Orton was too much.

That could very well be. At the time ATI and AMD were having talks, I'm sure that ATI senior management was able to hype up unprecedented bandwith (512 bit bus) on the R600, projections of 1GHz core clock, UVD on lower priced R6xx variants, etc. AMD bought into it hook, line, sinker, and totally underestimated competition from NVIDIA and Intel.

 

[jimmyjames123]

Yes, from a financial perspective. What I meant to do is justify why the acquisition will, in the end, have been very negative for ATI (in ways that perhaps couldn't have been predicted back then). The fact that the CPU division's bleeding has just begun will hurt ATI, and they'll be worse off than without the acquisition. Their position, relatively speaking, wouldn't have been so bad if they hadn't been acquired.

Yep, that's exactly right. The problem for ATI is that, since AMD is in so much debt and needs cash in a big way, then any cash generated by ATI will be quickly gobbled up by AMD. That is the disadvantage of having one company buy out another company. If the purchasing company starts to bleed cash, then they take cash from the purchased companies. I just don't see how AMD can afford to allocate hundreds of millions of dollars in developmental costs for GPU or GPU-specific items. So times will be very tough for ATI moving forward since it is doubtful they will be able to get the financial resources that they have been accustomed to in the past when developing new GPU's.

 

[jimmyjames123]

On a side note, Jen Hsun must be licking his chops. He would loveeeeeeee to get his hands on some of the ATI engineers If I was him, I'd be trying like hell to get the ATI engineers to come to the green team.

 

[Sxotty]

On a side note, Jen Hsun must be licking his chops. He would loveeeeeeee to get his hands on some of the ATI engineers If I was him, I'd be trying like hell to get the ATI engineers to come to the green team.

Do you really think so? It seems patents might be better at the moment. Nvidia has quite a few of both patents and engineers themselves though. But patents are not as hard to accommodate since you don't have to do anything when you ad more

 

[jimmyjames123]

Yeah, I'm sure Jen Hsun would love to have patent rights, but I'm pretty sure that he would be even more interested in getting ATI's top engineers on board. NVDA has a lot of engineers already, but they are growing as a company, and you can never have enough good engineering talent in Jen Hsun's view.

 

[Fox5]

Yeah, I'm sure Jen Hsun would love to have patent rights, but I'm pretty sure that he would be even more interested in getting ATI's top engineers on board. NVDA has a lot of engineers already, but they are growing as a company, and you can never have enough good engineering talent in Jen Hsun's view.

Maybe he'll want some cpu engineers so nvidia can completely develop its own platform. :devil's advocate:

 

[_xxx_]

I don't know the exact numbers since it's sadly not broken up like this, never was and never will I guess. However, from May 2005 to May 2006, you had tons of tape-outs/respins too, and marketing/dev costs too! I'm not sure what's magical about the R600 period that makes it more expensive than any other in ATI's history by a significant margin. Those costs were always there, they're not new.

The costs are higher than ever before, think die size, process, complexity. They never produced such an expensive chip before, nor a more complex/expensive card (power section, more layers etc.).

R520 period was probably seen by the public as a slip which won't be repeated that soon. The fact that it happened again and in an even worse manner with R600 sealed it all. R600 was their last straw concerning the leadership position in 3D gfx and they messed it up badly due to a overly ambitious design once again.

Yes, from a financial perspective. What I meant to do is justify why the acquisition will, in the end, have been very negative for ATI (in ways that perhaps couldn't have been predicted back then). The fact that the CPU division's bleeding has just begun will hurt ATI, and they'll be worse off than without the acquisition. Their position, relatively speaking, wouldn't have been so bad if they hadn't been acquired.

But would ATI still exist then? That's what I'd like to know. If the ~700 mio. value mentioned above is true, it wouldn't have been a nice ride for sure.

 

[_xxx_]

Maybe he'll want some cpu engineers so nvidia can completely develop its own platform. :devil's advocate:

I'm pretty sure they have been evaluating their options and maybe even some designs for quite a while already. Not saying that they will go that way, though. Hard to tell.

I don't think ATI would have been any more aggressive without the acquisition. Technically, they were in the same situation this way or the other - they had/have no more competitive design in the pipeline that would be doable in the same timespan.

 

[Anon Lamer]

I don't remember how bad their AGP design was, but I bet it had tons of issues too considering how god awful it was on MVP3.

Hehe on my KT133 board I can only run at agp 2x and no sidebanding with my 9700 pro otherwise I get crashes in civilization 4. It took a long time to find out why. Its the Via experience! Another VIA happening that was not so long ago was that their SATA controller corrupted disks that were too fast.

 

[silent_guy]

The costs are higher than ever before, think die size, process, complexity. They never produced such an expensive chip before, nor a more complex/expensive card (power section, more layers etc.).

You're exaggerating the relative cost of tape-outs. If it takes a conservative $300M to design a new architecture, a $1.5M full tape-out and even cheaper metal spins are not the kind of things that's going to sink a company...

Chip design expenses are heavily front loaded. A 6 month delay is painful, but for other reasons than overall project cost. Long before tape-out, a large part of designers is already working on other stuff and only called back temporarily when a bug is suspected on silicon.

If you're looking for reasons for bad financial performance, you should look at the lost opportunity cost and bad margins, not mundane stuff like the cost of a spin.

 

[pelly]

If you're looking for reasons for bad financial performance, you should look at the lost opportunity cost and bad margins, not mundane stuff like the cost of a spin.

I think people are merely saying that the "sum total" of a poorly-executed launch (or two) are to blame....Here's a small sampling of the issues at hand...

• Missing critical product cycles leads to:
  • Lower Market Share
  • Missed Revenue Targets
• Poor Margins lead to:
  • Missed Revenue Targets
• Late Arrival to Market "could" lead to:
  • Developers have less time with company A hardware, more time spent optimizing and coding with company B hardware
  • Less time with "final" silicon to internally iron out driver issues and optimize performance
  • Possibly missing key game releases leaving competitor B as the only option (no competition)
• Issues specifically with a flagship GPU typically lead to:
  • Consumer confidence in brand falters throughout entire product range
  • Investor confidence in brand falters
  • Poor publicity in media (due to losing benchmarks or not even competing in benchmarks due to lack of availability)
  • Lack of confidence in new process sizes/technologies (ie: NV30)

 

[_xxx_]

You're exaggerating the relative cost of tape-outs. If it takes a conservative $300M to design a new architecture, a $1.5M full tape-out and even cheaper metal spins are not the kind of things that's going to sink a company...

I'm talking about mass-production costs, not the tapeouts. These are evidently much higher for the R600 than say the G80, the card is more complex as well and some components are required to be higher grade of the more expensive sort too (power section for example).

Let's not even start with costs for a couple of (ass-uming here, otherwise why all the delays?) heftier redisigns which likely occured through R600 development.

 

[dizietsma]

Have you forgotten the infamous VIA pci bug in the 686b southbridge?

To forget something you have to remember it in the first place and this seems to be connected to large data transfers with Creative sound cards involved. Hardly a fatal flaw and VIA promised a BIOS upgrade to fix it. Hardly a big deal. Infamous ? Like a medium sized Mcdonalds vanilla milk shake bought on a Wednesday is infamous. Maybe so.

Whereas with the unlocked PCI bus problem with K8 chipsets there was no BIOS fix. Actually there was a promised BIOS fix that never materialised and then they brought out a new chipset to fix it. Meanwhile everyone, including myself, had gone to nvida......

 

[silent_guy]

I'm talking about mass-production costs, not the tapeouts. These are evidently much higher for the R600 than say the G80, the card is more complex as well and some components are required to be higher grade of the more expensive sort too (power section for example).

Oops. Yes, then I'm in perfect agreement.

Let's not even start with costs for a couple of (ass-uming here, otherwise why all the delays?) heftier redisigns which likely occured through R600 development.

If we assume that January or begin December was the intended release, then a hefty redesign is not very likely, since it doesn't look like they had to do a base spin. My current bet for the delay is that they suddenly needed to reposition their product (and redesign the PCB?) when it turned out to be non-competitive at the ultra high end. That, and drivers that were just not ready.

 

[chavvdarrr]

To forget something you have to remember it in the first place and this seems to be connected to large data transfers with Creative sound cards involved. Hardly a fatal flaw and VIA promised a BIOS upgrade to fix it. Hardly a big deal. Infamous ? Like a medium sized Mcdonalds vanilla milk shake bought on a Wednesday is infamous. Maybe so.

Whereas with the unlocked PCI bus problem with K8 chipsets there was no BIOS fix. Actually there was a promised BIOS fix that never materialised and then they brought out a new chipset to fix it. Meanwhile everyone, including myself, had gone to nvida......

Huh?
clicking sound when you listen mp3's + data corruption - "hardly a fatal flow" ?!
And then you state that unlocked PCi bus which relevant only to overclockers is The Sin ?

 

[BRiT]

To forget something you have to remember it in the first place and this seems to be connected to large data transfers with Creative sound cards involved. Hardly a fatal flaw and VIA promised a BIOS upgrade to fix it. Hardly a big deal. Infamous ?

It's worse than that. It took them over five months to come out with a bios fix. Many were affected to the point that they could not even install Windows XP. It always kept corrupting the drives. It didn't matter if you used RAID-0, RAID-1, single EIDE 7200 rpm or even a single 5400 rpm drive and yes, even without any Creative product installed in the system. It completely prevented the system from being stable at all. It didn't matter if you RMA'd the board 3 times, all boards were equally screwed. The only recourse was to return the board for a refund and move onto a board using a completely different chipset.

I suffered through that. It was the death kneel of VIA for me. You couldn't even visit any VIA discussion boards without running into masses of people being effected by the same very issue. It was infamous.

 

[Aerows]

Have you forgotten the infamous VIA pci bug in the 686b southbridge?

Gollom voice

:::AAAYYY!!! It hurts us, it hurts us! Stop reminding it hurts us!!:::

In other words, no, many have not forgotten

 

[Anon Lamer]

How different would AMDs situation be today if they had bought SiS instead of ATI?