Official AMD FX Bulldozer review thread!

[uOpt]

Originally Posted by Rollo

I don't buy into the whole "If AMD goes, intel and NVIDIA will put the screws to us!" position.

intel and NVIDIA need us to keep buying their products on a regular basis to keep their doors open, so they can't just say "Haha, you are all screwed. Our new price points are $1000,$2000, and $3000, pay it or do without.".

The number of people who would pay that would be too small to cover their overhead. (especially in a world wide recession)

Yeah but they can still shut out the door unlocked multiplier or all overclocking for that matter.

And they can make wanting ECC really expensive and inconvenient. Have you tried assembling a single-socket 1155 Xeon system with ECC? Seen what kind of ridiculous offers there are in the mainboards section? How few slots? In $200 boards?

Of course cheap unlocked multipliers are a new thing even for AMD but their move got Intel rolling, too.

[Rollo]

Originally Posted by uOpt

Yeah but they can still shut out the door unlocked multiplier or all overclocking for that matter.

And they can make wanting ECC really expensive and inconvenient. Have you tried assembling a single-socket 1155 Xeon system with ECC? Seen what kind of ridiculous offers there are in the mainboards section? How few slots? In $200 boards?

Of course cheap unlocked multipliers are a new thing even for AMD but their move got Intel rolling, too.

Actually I think selling CPUs in a stale market is what got intel to do things like unlocked multipliers.

They have 80% of the market, do you think they actually were sitting around the R&D dept saying "Willikers guys! AMD is selling 20 of every 100 CPUs these days, we'd better follow suit on the unlocked multipler thing before they claim 21%!".

SB cpus are so much better than PhenomIIs they could have sold just as many with a locked multiplier.

[uOpt]

Originally Posted by Rollo

Actually I think selling CPUs in a stale market is what got intel to do things like unlocked multipliers.

They have 80% of the market, do you think they actually were sitting around the R&D dept saying "Willikers guys! AMD is selling 20 of every 100 CPUs these days, we'd better follow suit on the unlocked multipler thing before they claim 21%!".

SB cpus are so much better than PhenomIIs they could have sold just as many with a locked multiplier.

Your numbers are all off but still you are right. That's exactly why Intel started offering unlocked multipliers.

[Rollo]

Originally Posted by uOpt

Your numbers are all off but still you are right. That's exactly why Intel started offering unlocked multipliers.

If AMD had something comparable and they offered unlocked and intel did not, I could see intel thinking they better offer unlocked. As there was no parity in performance, intel had no reason other than trying to tweak the market they already owned.

AMD just hasn't been a force of comeptition since Core2 launched. Competition can keep prices down and drive innovation, but not if it's weaksauce competition.


I think buying ATi crippled AMD to an extent they just could not recover.

[Viral]

I don't. They wouldn't have anything to differentiate them from Intel if they didn't have ATI.

Time's have changed, we're very unlikely to see another Athlon 64. Expect more along the lines of the original Athlon vs P3 for the best AMD can do on the CPU side these days. Intel dropped the ball with the original P4 and dropped it again by betting on Prescott to raise frequencies. This coupled with the fact that they weren't in the compiler game back then means AMD will probably never have a CPU like the t-bird or Athlon64/x2 again. Luckily they have much better GPU's than intel, that's the only card they have left to play.

[Viral]

Originally Posted by Rollo

And yet which company made more money this year, is worth about 8X as much, could buy the other with cash, likely won't be announcing their departure from high end gaming this week, and isn't laying off all their best people while the world wonders if they'll be in business next year?

That doesn't mean AMD's strategy was bad in the way you said it was. You completely ignored the main focus of it.

Point is AMD's strategy worked well for the gaming crowd. Unfortunately there's not much money to be had there compared to other, more lucrative markets. That's where NV cashed in. Now if AMD could only get their software support up to scratch they still might have a good chance with this strategy, although it's still unclear what die sizes GCN is targeting - I think the moderate die vs large die at 5-10% performance penalty is a worth while strategy though.

[Rollo]

Originally Posted by Viral

That doesn't mean AMD's strategy was bad in the way you said it was. You completely ignored the main focus of it.

Actually what I said still applies. Yes the ratio of die size to prices was roughly equivalent, but what if they had split the difference and pocketed another $75/chip? Once you get people expecting to pay much less for your products, hard to go back to higher prices.

Or are you saying AMD didn't need the money?

[Viral]

Originally Posted by Rollo

Actually what I said still applies. Yes the ratio of die size to prices was roughly equivalent, but what if they had split the difference and pocketed another $75/chip? Once you get people expecting to pay much less for your products, hard to go back to higher prices.

Or are you saying AMD didn't need the money?

This is what you said,

All the "their strategy is to make second place GPUs and sell them cheap!" PR spin in the world can't change these facts. Like I posted long ago, when ATi lowered their prices and started selling their top end products for $300 instead of $500 and their midrange for $200 it was the LAST thing they needed. They were buying market share at the cost of profit per unit (same as AMD) at a time when they needed money.

Now you're admitting they didn't sacrifice profit per unit. Flip-Flop.

Point is that last 5-10% extra performance that NV are getting comes at the cost of 30-40% more transistors. There is a market for that, and that market will happily pay AMD for its products if they return to that strategy and deliver on the performance front. It's a tough strategy though - having the halo product is great and all (although AMD technically still have that) but it's getting harder and harder as we move on and hit power and heat barriers to deliver these ultra high end products. Then you have to either bin these products for the slightly lower end or spend more on R&D for a mid-high end product. AMD's strategy was very valid on the desktop front IMHO, for other markets, maybe they did sacrifice? They never really had a foot hold in these markets in the first place, however.

[Rollo]

It's likely all a moot point. I read there is one GPU PR guy left in Markham, and when you add that to the layoffs in GPU engineering at high levels it doesn't really sound like the future is GPUs does it?

[Rollo]

XBit reviewed the Bulldozer again this week:

http://www.xbitlabs.com/articles/cpu...6100-4100.html

In other Bulldozer news, MS releases BeefDozer patch, retracts it promptly.

[T.h.o.m.a.s.]

the reality check tells otherwise :s
i wonder if they would do this check amongst users of this forum what
the outcome would be

[Xion X2]

Discrete CPUs are no longer AMD's focus. That's why you have the KillDozer epic failure. They began work on this before they acquired ATi.

AMD's acquisition of ATI was so they could create Fusion. That's their baby and their main focus. It's been in their plans all along to merge the CPU with the GPU for an APU as that seems to be the direction things are moving in. Intel's also incorporating this strategy, although to a lesser extent.

I think discrete CPUs will be a thing of the past 4-5 years from now as even low-end laptops need graphics for today's tasks. What will be interesting to see is how well Intel can keep pace (or possibly outpace) AMD in the APU market considering there are better graphics engineers in-house at AMD through the ATI acquisition than are at Intel (or would seem to be at Intel.)

What we're looking at now is a milestone transition in the industry. You have Intel, who's specialized in discrete CPUs. Then you have AMD, who's specialized in both discrete CPUs and in GPUs through the ATI acquisition. Then you have Nvidia, who's specialized in discrete GPUs and now moving into CPUs through the mobile space.

Should be fun to watch play out.