What if bulldozer...

[MisterMac]

....had come out in Q2 2011?

Or late Q1?



Same performance, but actually on track with SB?



How would the AM3+ socket owners feel? would it be more acceptable and we could put focus on AMD rushing it and accepting a revision maybe.. now that would have fixed it?


Assuming they could deliver the same product in ´feb/march/april, would the disappiontment be any less to you?


It seems to me, it failed but even so more cause of the timing now being completely off.

They're playing catch up and have had to cancel a planned iteration of BD(Was it komodo? i forget).
Making users more screwed and showing that they're in serious trouble, was this a "match/game" and simply are trailing far behind.

 

[Red Storm]

BD would still be a disappointment. It's just not designed for the consumer desktop market. Hopefully it makes some money for them in the server market.

I don't see how one product's delays are going to cause delays for future products that are likely already in the design process (and by a different team too).

The sad thing is that AMD will always be playing catch up with Intel (even when they had the better processor...). There's no getting around the huge advantages Intel has over them.

 

[blckgrffn]

If it had come out when Intel was recalling a bunch of mobo's and taking a $1 billion write down to cover it, it would have at least been spared some of the nerd rage that was at that point being directed at Intel. That would have been in some way beneficial to AMD but that would be hard to measure, I think.

 

[sm625]

BD could have been released in 2010 and it would still not have gotten glowing reviews.

 

[LoneNinja]

Bulldozer is far to power hungry, I wouldn't use it if given to me for free right now. I really hope this is an issue with manufacturing and yields right now, hopefully they can get voltages/power consumption down.

 

[Martimus]

Bulldozer's biggest problem is that it was designed to go much faster than it is. Due to the manufacturing problems at the foundry, they just can't get those speeds. If there were no issues with the process, then it would have been released on time in all likelyhood based on what has been said, but alas that did not happen.

The good news about that is that we should see relatively large increases in clock speed as the 32nm process becomes more mature. So we should see better chips down the road. Since Intel has quite a bit of head room in their design already, they will likely respond with similar increases in clock speed as well, so we have a pretty nice win-win scenario for us as consumers. (although since Intel should be able to match any clock increases that AMD produces through the near future, that may preclude AMD from even attempting any clock speed increases and instead just use increased yields to help lower power consumption and reduce costs on packaged heat sinks.)

 

[86waterpumper]

huh? There are not going to be BIG speed increases...even most intel 2500 and 2600k chips only o/c to 4.5ghz across the board on normal voltages. Yeah some do more but you see most people running them right around 4.5. The bd already turbos at least to 4ghz does it not?

 

[Martimus]

huh? There are not going to be BIG speed increases...even most intel 2500 and 2600k chips only o/c to 4.5ghz across the board on normal voltages. Yeah some do more but you see most people running them right around 4.5. The bd already turbos at least to 4ghz does it not?

The BD was designed to run at a stock speed of 4.3GHz (Current Thuban speed plus 30%). So it is not running at it's designed clock speed.

For an architecture that was designed to scale to clock speeds 30% higher than its predecessor, Bulldozer doesn't seem to be coming anywhere close.

http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested

 

[IntelUser2000]

The BD was designed to run at a stock speed of 4.3GHz (Current Thuban speed plus 30&#37. So it is not running at it's designed clock speed.

Pardon me for saying, but I think the increased clock speed was merely to offset the decrease in performance in clock, or at least a buffer if it failed. If they did not do it, and delivered 5% better IPC, it would have been worse. Therein lies the problem. I don't think even using Intel's process they can go much better than 5GHz.

Prescott wasn't running at designed clock speeds either. But I'm not sure if its possible to even reach 10GHz on a practical chip if you need to go all out on everything to reach world record 8GHz speeds today.

In a time it takes to complete a cycle, light in a vacuum can only travel 60mm on a 5GHz chip. That's really not a lot considering there are far more limiting factors like wire resistance, etc. 8-9GHz might not be far off from the practical limit, forever.

 

[cytg111]

the sad part is really that we will see ivey pushed back yet another 6 months (we'll see it in a year from now tops), and haswell .. well.. haswell has-it-well in the diatant future.. it will be the p2/3-500mhz days all over again... my next cpu upgrade will be SB or ivey, and problary a dual socket mobo.. cause the architecture is going to last for a looong time.

 

[IntelUser2000]

the sad part is really that we will see ivey pushed back yet another 6 months (we'll see it in a year from now tops), and haswell

I can't speak for future chips, but there's no reason at all for Intel to purposely delay their first 22nm chip, Ivy Bridge. The fabs are making chips now, the only thing they need is to get enough volume and finishing on validation. Any unnecessary delay means they are missing out on lower costs from going to 22nm, while the fabs are idling.

There's another problem. Delays in process tech introductions carry onto the subsequent nodes, and that cuts into Intel's process tech lead over the competition. They may be able to afford Ivy Bridge being delayed, but not process tech, when they'll need that for low power chips.

Prior to 90nm, they were on the summer release schedule for process tech introductions. Now we are 4-6 months later at Q4/Q1, and have been for 4 generations.

 

[86waterpumper]

The BD was designed to run at a stock speed of 4.3GHz (Current Thuban speed plus 30&#37. So it is not running at it's designed clock speed.

How is it not running close to the designed clock speed when it turbos close to that speed already? My 2500k runs at 1600mhz some of the time, but I still say it's o/c to 4.2 ghz etc so I"m not following you. Are you saying the bd was designed to have a stock freq. of 4.3 and turbo up to 5ghz or something? I find that tough to swallow. Do you really see them getting the yields so good that they can increase a entire ghz in speed?

 

[Maximilian]

I dont even think BD would've fared that well if it was released in 2009.

 

[bryanW1995]

Q2 2011 was probably its best possible release window between jan 2010 and oct 2011. As mentioned, there was enough nerd rage against intel that AMD could have avoided quite a bit of bad press and could have pointed the finger at intel and said "See, we'll eventually get it right. Even intel makes mistakes!" When they couldn't get it out by the end of Q2 I feared the worst, and things have panned out even worse than the worst. Is this worster? The worstest? Maybe we should just coin a new term: AMD pulled a Hector.

 

[sangyup81]

Then I hope they would have tweaked it by now

 

[Lonbjerg]

NetBurst's biggest problem is that it was designed to go much faster than it is.*snip*

FYP...ironic isn't it?

 

[hooflung]

BD would still be a disappointment. It's just not designed for the consumer desktop market.

Last time I checked the consumer desktop market was in shambles and margins were already razor thin.

 

[bryanW1995]

Really? I wasn't aware that intel's 99.9999 % profit margin (estimated) was due to their ginormous profits on ssd's. /sarcasm

 

[Nemesis 1]

BD would still be a disappointment. It's just not designed for the consumer desktop market. Hopefully it makes some money for them in the server market.

I don't see how one product's delays are going to cause delays for future products that are likely already in the design process (and by a different team too).

The sad thing is that AMD will always be playing catch up with Intel (even when they had the better processor...). There's no getting around the huge advantages Intel has over them.

AMD has what six% of the server market . I suspect next Qt that will be 5% or less . Which brings up another problem AMD has . Whos going to recompile for BD with a 6% server share and shrinking daily . NO ONE.

 

[Nemesis 1]

Last time I checked the consumer desktop market was in shambles and margins were already razor thin.

I agree it must be hard for intel to suffer threw 60% margins . Its not a growing market, that doesn't mean a whole lot at this stage of the game.

 

[cebalrai]

BD could have been released in 2010 and it would still not have gotten glowing reviews.

Why not? It's far more competitive against Nehalem and it wouldn't have been stacked up against AMD's own 1100T.

 

[Maximilian]

Why not? It's far more competitive against Nehalem and it wouldn't have been stacked up against AMD's own 1100T.

It cant beat a stock i7 920 in most things, its lousy vs nehalem and this is at 32nm. Back then if it was released it would have been 45nm making for an even more monumental failure.

 

[cebalrai]

It cant beat a stock i7 920 in most things, its lousy vs nehalem and this is at 32nm. Back then if it was released it would have been 45nm making for an even more monumental failure.

No it wouldn't have AFAIK.

 

[soccerballtux]

Bulldozer's biggest problem is that it was designed to go much faster than it is. Due to the manufacturing problems at the foundry, they just can't get those speeds. If there were no issues with the process, then it would have been released on time in all likelyhood based on what has been said, but alas that did not happen.

The good news about that is that we should see relatively large increases in clock speed as the 32nm process becomes more mature. So we should see better chips down the road. Since Intel has quite a bit of head room in their design already, they will likely respond with similar increases in clock speed as well, so we have a pretty nice win-win scenario for us as consumers. (although since Intel should be able to match any clock increases that AMD produces through the near future, that may preclude AMD from even attempting any clock speed increases and instead just use increased yields to help lower power consumption and reduce costs on packaged heat sinks.)

how fast was it supposed to go, 5+ghz? Longer pipeline should have let it no?

 

[IlllI]

if bd had come out much earlier we might be discussing bd2 by now.