Bulldozer TurboCore 2.0

sirmo

Gawd
Joined
Sep 23, 2009
Messages
516
It looks like AMD's TDP based Turbo is very aggressive.

AMD FX-8150 (max turbo 4.2Ghz)

  1. Two heavily loaded cores: 4.2Ghz
  2. 3-4 heavily loaded cores: 3.6-3.9Ghz
  3. 8 loaded cores: 3.6-3.9Ghz
  4. 8 heavily loaded cores: 3.6Ghz

http://www.youtube.com/watch?v=Aale7YSvhJk

Compare this to Sandy Bridge:

When running a single threaded application, the Core i7-2600K will power gate three of its four cores and turbo the fourth core as high as 3.8GHz. Even with two cores active, the 32nm chip can run them both up to 3.7GHz.
 
Last edited:
That should make it perform pretty well in non-enthusiasts rigs.

I imagine most of us here will turn turbocore off in order to get max stable overclocks anyway, so it is probably less interesting for us.
 
Pentium IV had agresive clockspeeds too ;)

Speaking from experience there buddy hehe.

Zarathustra[H];1037803413 said:
That should make it perform pretty well in non-enthusiasts rigs.

I imagine most of us here will turn turbocore off in order to get max stable overclocks anyway, so it is probably less interesting for us.

Yeah we will most likely turn that off. Still I want to test it out myself, when BD has been out awile and someone like freidns or family members ask what they should buy or they want games to run the way we overclockers run them we can recommend them a BD if they're interested in a cpu with alot of cores and power savings, with the best pricing.

You have to do that or you get to many Intel loving posters like the ones above hehe.

And we really are not sure yet if the more stable overclocks will be found the traditional way or with AMD BD newer power design. ;)
 
I hope they make that configurable.

It's nice to have the high clocks when you need them, My wifes SB rig runs at 1.6Ghz in windows until a game is launched and then kicks up to 4.7Ghz. I'd like to do the same with BD.
 
Will all this OCing jazz, I wonder if they'll have a ITX capable of OCing because the Z68s so far in that factor suck a big one.
 
I wonder if they'll have a ITX capable of OCing because the Z68s so far in that factor suck a big one.

I do not think this is at all where BD will shine. I mean low power small form factor overclocking.
 
The 8150 will do that at 5 or more threaded applications for sure. For less than 5 threads we will have to see how much more frequency it will need over the i5 to be equal.

All it has to beat are 4.5ghz 2500Ks.

Are you trying to do that in the itx form factor? Maybe the 2 module version at 5.X GHz so to keep the power to a more reasonable level.
 
I hope they make that configurable.

It's nice to have the high clocks when you need them, My wifes SB rig runs at 1.6Ghz in windows until a game is launched and then kicks up to 4.7Ghz. I'd like to do the same with BD.

It should be configurable in bios, on/off. When it is on, it will not kick in unless your apps need it, so it won't push the clocks up (more power and heat) when you are doing email, it will only kick in when you need the extra speed.
 
Probably won't, even at 5Ghz. If it can get close for less that is still fine though.
 
who cares about cycles per second.

heck a heater prob runs at higher cycles.


what we need is how fast can it do instructions per cycle and how quick is a single cycle.
 
who cares about cycles per second.

heck a heater prob runs at higher cycles.


what we need is how fast can it do instructions per cycle and how quick is a single cycle.

That's true. Maybe it will blow that SB heater you speak of out of the water huh? :p
 
who cares about cycles per second.

heck a heater prob runs at higher cycles.


what we need is how fast can it do instructions per cycle and how quick is a single cycle.

Cycles per second determine the performance as much as the IPC does.

CyclesPerSecond*IPC = Performance

If BD delivers 90% of the IPC but 120% of the cycles per second, it's a clear winner.

Or look at it this way. Even though 5870 had less "IPC" but more clock it still gave Fermi run for its money, while at the same time being more power efficient.
 
Precisely. Too many people get caught up in IPC. To turn up the contrast a bit, what if IPC went up by 2X but the clock speed was only 800MHz?

You have to be careful when picking a single data point. When you use a single data point you risk making bad decisions.
 
If BD delivers 90% of the IPC but 120% of the cycles per second, it's a clear winner.

Agreed, however. I would be shocked if either of those were true let alone both. I would expect more like 85% of the IPC and 110% of the frequency at 125% of the power usage. However this does not take into account of the 6 (maybe 5)+ thread advantage of BD.
 
Last edited:
Precisely. Too many people get caught up in IPC. To turn up the contrast a bit, what if IPC went up by 2X but the clock speed was only 800MHz?

You have to be careful when picking a single data point. When you use a single data point you risk making bad decisions.

True, both factors need to be weighed equally.

That being said, an arch with a low IPC can only remain competitive if you truly have enough clock headroom to compensate for it.

Seeing that current Sandy Bridge can hit 4.7 on air means that Intel probably isn't too far off from having good enough yields to crank up the clock speed within their desire thermal envelope.

I can't help but think if a 2600k and a FX8150 were to battle it out air cooled, with or without thermal envelope in mind Intels max clock isn't going to be significance different than that of bulldozer. So with both topping out at about the same clock, the chip with the higher IPC will take the crown.

In order to make low IPC work you have to be able to school the competition in clock frequency, and without liquid nitrogen it doesn't look like BD is doing it unfortunately :(

I do - however - applaud the well designed turbo core on BD. It lOoks
Like it will help non overclockers a lot.

The problem is that even while at a turbo of 4.2Ghz, unless there is a real IPC improvement over K10 even at 4.2Ghz a BD core will just match A 2.6ghz i3-2120T core...

I hate to be a downer. I REALLY want BD to be a killer, but no data or theoriIng I can see supports this. :(
 
Last edited:
Zarathustra[H];1037805392 said:
True, both factors need to be weighed equally.

That being said, an arch with a low IPC can only remain competitive if you truly have enough clock headroom to compensate for it.

Seeing that current Sandy Bridge can hit 4.7 on air means that Intel probably isn't too far off from having good enough yields to crank up the clock speed within their desire thermal envelope.

I can't help but think if a 2600k and a FX8150 were to battle it out air cooled, with or without thermal envelope in mind Intels max clock isn't going to be significance different than that of bulldozer. So with both topping out at about the same clock, the chip with the higher IPC will take the crown.

In order to make low IPC work you have to be able to school the competition in clock frequency, and without liquid nitrogen it doesn't look like BD is doing it unfortunately :(
(

Except one thing is missing in your argument. And that's BD's longer pipeline, which allows for higher clocks on the similar fab process, as demonstrated by "well above 5Ghz OC on Air" with eng. samples.

I think we can all blame Netburst's failure for the prevailing opinion: "longer pipeline is bad" and "IPC is the only thing that matters" misconception. Netburst had a lot of things wrong with it.

Here is an interesting read: http://abinstein.blogspot.com/2010/09/ipc-myths.html

To summarize the important point for the lazy:

IPC vs. higher clocks (longer pipeline) is a tradeoff. Like any tradeoff relation, you don't get optimal results by sliding towards either edge.
 
Last edited:
It seems like some people just cannot be realistic about Bulldozer. They're probably the same people that were wholly unrealistic about Phenom, too.

So let's take some of the shady-looking numbers floating around. Assuming BD delivers 90% of the IPC but 120% of the clock speed (which is what somebody said would make it a winner), and comparing it to a 4.5GhZ SB CPU (which is apparently what you guys think it needs to beat)...

Bulldozer will beat a 4.5GhZ Core i5 2500K when it is clocked to 5.4GhZ.

Now, let's put it up against all of those of us who are in the 5GhZ+ club with 2600Ks. You'd need BD clocked at 6GhZ. How many people are going to phase-change their BD chips to try and beat Intel's high-end MAINSTREAM processors that are performing that well on extreme air/standard water cooling?

Ridiculously fuzzy math, but I think you get the point.

Also, since Teletran thinks the SB chips make good heaters or something... mine doesn't really seem to heat the room up at all. Even bitcoin mining with 4 CPU threads and 2 GPU threads on my 6870s only heats the room up about 3 degrees in the course of 4 hours. My iguana loves it.
 
It seems like some people just cannot be realistic about Bulldozer. They're probably the same people that were wholly unrealistic about Phenom, too.

So let's take some of the shady-looking numbers floating around. Assuming BD delivers 90% of the IPC but 120% of the clock speed (which is what somebody said would make it a winner), and comparing it to a 4.5GhZ SB CPU (which is apparently what you guys think it needs to beat)...

Bulldozer will beat a 4.5GhZ Core i5 2500K when it is clocked to 5.4GhZ.

Now, let's put it up against all of those of us who are in the 5GhZ+ club with 2600Ks. You'd need BD clocked at 6GhZ. How many people are going to phase-change their BD chips to try and beat Intel's high-end MAINSTREAM processors that are performing that well on extreme air/standard water cooling?

Ridiculously fuzzy math, but I think you get the point.

Also, since Teletran thinks the SB chips make good heaters or something... mine doesn't really seem to heat the room up at all. Even bitcoin mining with 4 CPU threads and 2 GPU threads on my 6870s only heats the room up about 3 degrees in the course of 4 hours. My iguana loves it.

PCperspective guys witnessed BD eng. samples go well above 5Ghz on just air, all 8 cores enabled (at the 8.4Ghz world record breaking event). If eng. samples on a brand new 32nm can do well above 5Ghz on air, I think 6Ghz on quality water may be doable with production silicon. Also we're talking single thread performance, factor in 8 cores and you got yourself some serious computing power.

Also this is the AMD section of the forums, how dare I be optimistic?
 
It seems like some people just cannot be realistic about Bulldozer. They're probably the same people that were wholly unrealistic about Phenom, too.

So let's take some of the shady-looking numbers floating around. Assuming BD delivers 90% of the IPC but 120% of the clock speed (which is what somebody said would make it a winner), and comparing it to a 4.5GhZ SB CPU (which is apparently what you guys think it needs to beat)...

Bulldozer will beat a 4.5GhZ Core i5 2500K when it is clocked to 5.4GhZ.

Now, let's put it up against all of those of us who are in the 5GhZ+ club with 2600Ks. You'd need BD clocked at 6GhZ. How many people are going to phase-change their BD chips to try and beat Intel's high-end MAINSTREAM processors that are performing that well on extreme air/standard water cooling?

Ridiculously fuzzy math, but I think you get the point.

Also, since Teletran thinks the SB chips make good heaters or something... mine doesn't really seem to heat the room up at all. Even bitcoin mining with 4 CPU threads and 2 GPU threads on my 6870s only heats the room up about 3 degrees in the course of 4 hours. My iguana loves it.

I think you're the one who needs to pay attention.

He just threw out 90% IPC and 120% as example numbers. Nowhere did he say those were indicative of the actual numbers. He didn't say that's what they were doing, he was trying to point out that IPC wasn't everything, like some people are so focused on.

Also, very rarely does Sandy Bridge hit 5 ghz without a good watercooling setup. 4.5 ghz is the typical 24/7 overclock of most Sandy Bridge users. If saying, for example, Bulldozer is able to achieve more than 5 ghz consistently on air cooling, which is what some of the rumors are stating, then it'll be as good as Sandy Bridge. And if it can hit 5+ ghz consistently on air, it's not too much of a stretch for it to hit at least 5.5 ghz on water.

By this point, most of us are not looking for clear winners over the 2600k or 2500k. We're looking for performance at least equivalent to the 2500k, if not the 2600k.

BTW, you need to do a bit of math yourself. .9 x 1.2 = 1.08, which means 8% more performance than Sandy Bridge. To make it equivalent, Bulldozer would need to be clocked at 5 ghz to match a 4.5 ghz Sandy Bridge, and 5.55 ghz to match a 5 ghz Sandy Bridge.
 
As I stated, I was just using those theoretical numbers so people could see that it isn't a small difference that's going to square them up. And yes, I'm well aware that 90% of 120 is 108.

There's a very large 5GhZ+ club amongst 2600K owners (I'm one of them, with hyper-threading on, no cheating). Mine hits 5.1GhZ and is stable enough for benchmarking over the course of 2-3 hours well within thermal limits of the chip using just a Corsair H50 and MX-2 (and on about 1.48v which is what many SB OCers believe to be entirely sane). Remember, though, these Bulldozer chips are AMD's high-performance parts and Sandy Bridge is Intel's mainstream parts.

Not hating on AMD, just being more realistic than the super hardcore fanboys. My Phenom II is enjoying home media center/virtualized test lab life in the living room and my K6-2 chip is proudly tucked away for showing friends how [H] I am lol
 
As I stated, I was just using those theoretical numbers so people could see that it isn't a small difference that's going to square them up. And yes, I'm well aware that 90% of 120 is 108.

There's a very large 5GhZ+ club amongst 2600K owners (I'm one of them, with hyper-threading on, no cheating). Mine hits 5.1GhZ and is stable enough for benchmarking over the course of 2-3 hours well within thermal limits of the chip using just a Corsair H50 and MX-2 (and on about 1.48v which is what many SB OCers believe to be entirely sane). Remember, though, these Bulldozer chips are AMD's high-performance parts and Sandy Bridge is Intel's mainstream parts.

Not hating on AMD, just being more realistic than the super hardcore fanboys. My Phenom II is enjoying home media center/virtualized test lab life in the living room and my K6-2 chip is proudly tucked away for showing friends how [H] I am lol

The [H] club is nowhere near indicative of the general population. Actually, any tech club is not indicative of the general population. As I said before, 4.5 ghz is the typical 24/7 overclock of Sandy Bridge owners. Not 4.8 and not 5.1 (even you don't run it at 5.1). You want to compare bench runs? Bulldozer gets up to 8.4 ghz on liquid helium, Sandy Bridge can't even begin to compare against that.

We are all very well aware of the gap that Bulldozer has to close. However, what irked me was you taking one's person hypothetical example to illustrate a point, and (effectively) saying that he was believing that was fact, and implying that he was stupid for thinking so.

And actually, most people don't even expect Sandy Bridge performance. Most people (like me) just want to see it at least match Nehalem, so that AMD actually has a strong competitor. Coming close to Sandy Bridge is a big plus, and would be considered a win for AMD. Being better than Sandy Bridge would be like a grand slam home run.
 
Determined by what? The price?

No, determined by Intel having socket 2011 to pick up where socket 1366 left off and leaving 1155 to pick up where 1156 left off.

Believe you me, sir, when Phenom II launched I said to myself "Intel is ridiculously expensive" but here I found myself realizing in June that BD wasn't coming anytime soon and an extra $600 to burn on PC upgrades was in my pocket. I got what I paid for, more than doubled my CPU performance of my PII system and paid less than double to do it (PII CPU+mobo ran me about $325, SB cpu+mobo ran me $520). For budget enthusiasts AMD will definitely be the choice, and the numbers are indicating it'll be between Nehalem and Sandy Bridge so it'll be the better bang for the buck. Better overall bang? Not likely.
 
It seems like some people just cannot be realistic about Bulldozer. They're probably the same people that were wholly unrealistic about Phenom, too.

So let's take some of the shady-looking numbers floating around. Assuming BD delivers 90% of the IPC but 120% of the clock speed (which is what somebody said would make it a winner), and comparing it to a 4.5GhZ SB CPU (which is apparently what you guys think it needs to beat)...

Bulldozer will beat a 4.5GhZ Core i5 2500K when it is clocked to 5.4GhZ.

Now, let's put it up against all of those of us who are in the 5GhZ+ club with 2600Ks. You'd need BD clocked at 6GhZ. How many people are going to phase-change their BD chips to try and beat Intel's high-end MAINSTREAM processors that are performing that well on extreme air/standard water cooling?

Ridiculously fuzzy math, but I think you get the point.

Also, since Teletran thinks the SB chips make good heaters or something... mine doesn't really seem to heat the room up at all. Even bitcoin mining with 4 CPU threads and 2 GPU threads on my 6870s only heats the room up about 3 degrees in the course of 4 hours. My iguana loves it.

Wouldn't that be 4950 mhz to match a 4.5ghz bd?
 
No, determined by Intel having socket 2011 to pick up where socket 1366 left off and leaving 1155 to pick up where 1156 left off.

Believe you me, sir, when Phenom II launched I said to myself "Intel is ridiculously expensive" but here I found myself realizing in June that BD wasn't coming anytime soon and an extra $600 to burn on PC upgrades was in my pocket. I got what I paid for, more than doubled my CPU performance of my PII system and paid less than double to do it (PII CPU+mobo ran me about $325, SB cpu+mobo ran me $520). For budget enthusiasts AMD will definitely be the choice, and the numbers are indicating it'll be between Nehalem and Sandy Bridge so it'll be the better bang for the buck. Better overall bang? Not likely.

What...?

AMD never really had anything to compete with 1366. AM3/Phenom II competed with 1156. AM3+/Phenom III will compete with 1155. AMD will not have anything that will compete with 2011 at the moment, and probably not in the near future. Just like they didn't have anything to compete with 1366 at all.

And for most consumers, bang for buck is everything. Especially when performance is close, and won't make a difference in real world computing. You want overall bang? Go for an EVGA SR-2 with hexcore Xeons. Or wait for the quad-2011 socket and get 4 8-core Sandy Bridge-E processors.
 
Precisely. Too many people get caught up in IPC. To turn up the contrast a bit, what if IPC went up by 2X but the clock speed was only 800MHz?

You have to be careful when picking a single data point. When you use a single data point you risk making bad decisions.

And you twist the IPC argument as always exagerating it to absurd.

We are enthusiast here so most of us realise that within same technological process one architecture will rarely have much more frequency headroom than the other unless the IPC difference is drastic like in PIV Prescotts vs AMD A64.

Perhaps for average joe you will be able to get similar performance to i5 by binning your cpus for higher clock speeds like you guys did with X4 vs i5 750 but the effect was we had something like 600 Mhz frequency reserve on Phenom II and 1Ghz+ on i5 which resulted in bad spanking at overclocked settings.
 
What...?

AMD never really had anything to compete with 1366. AM3/Phenom II competed with 1156. AM3+/Phenom III will compete with 1155. AMD will not have anything that will compete with 2011 at the moment, and probably not in the near future. Just like they didn't have anything to compete with 1366 at all.

Disagree. AMD had nothing to compete with either 1366 or 1156.

AM2/AM3/Phenom II competed with socket 775 solutions, and even at that, it has competed poorly since the launch of the Core 2 Duo.

Hopefully this BD launch will bring them back in the game, but I'm doubtful.
 
And you twist the IPC argument as always exagerating it to absurd.

We are enthusiast here so most of us realise that within same technological process one architecture will rarely have much more frequency headroom than the other unless the IPC difference is drastic like in PIV Prescotts vs AMD A64.

Perhaps for average joe you will be able to get similar performance to i5 by binning your cpus for higher clock speeds like you guys did with X4 vs i5 750 but the effect was we had something like 600 Mhz frequency reserve on Phenom II and 1Ghz+ on i5 which resulted in bad spanking at overclocked settings.

Your entire argument falls apart when you learn that: BD has 2.4 Ghz reserve on SB and it's not even out yet. SB OC record is 6Ghz, BD's OC record is a world record @ 8.4Ghz.
 
Your entire argument falls apart when you learn that: BD has 2.4 Ghz reserve on SB and it's not even out yet. SB OC record is 6Ghz, BD's OC record is a world record @ 8.4Ghz.

Records on liquid nitrogen or other extreme cooling don't mean shit.

To be a better predictor of what might actually get released based on thermal envelopes we should look at OC records on stock boxed coolers :p
 
Zarathustra[H];1037806880 said:
Records on liquid nitrogen or other extreme cooling don't mean shit.

To be a better predictor of what might actually get released based on thermal envelopes we should look at OC records on stock boxed coolers :p

Yeah because we want to compare and know who ships a better stock cooler. If from the world record you can't derive that BD is going to clock higher than SB, than I don't know what to say.
 
If from the world record you can't derive that BD is going to clock higher than SB, than I don't know what to say.

I can't. This world record is meaningless in the real world. All I can derive from the world record is that under extreme cooling and extreme voltage and no load the BD rocks when you disable 6 of its 8 cores.
 
Yeah because we want to compare and know who ships a better stock cooler. If from the world record you can't derive that BD is going to clock higher than SB, than I don't know what to say.

Again, the world record doesn't mean anything.

If we exclude overclocking for a moment, in order for it to be a viable product it can't exceed a TDP of 125W. If AMD had a BD today that could clock higher than the planned FX 8150 and still not exceed 125W TDP then they would be releasing it, to make more of a challenge to Intel's lead. They don't, and that's why they have been forced to develop a smarter Turbocore system, but this is really more of a bandaid than anything else.

Now if we include overclocking, BD may be great for the few people in the community on phase change or chilled water solutions, but that's not very many of us. In the overclocking video AMD showed that their cherry picked BD sample was able to hit 4.7Ghz on a sealed water cooling solution. That's about the same clock speed as you get out of a Sandy Bridge with similar, but SB may have as much as a 50% IPC lead (if BD doesn't improve significantly on K10).

How well each architecture performs on extreme cooling like LN2 or Helium is utterly irrelevant today.

It may be important down the line once we see process maturity. As I recall Phenom II - when launched - would hit 4.2Ghz or thereabouts on cherry picked samples with extreme cooling when first launched. Now people are doing that on air, because of GlobalFoundries ridiculously mature 45nm process.

Hopefully - however - AMD has learned their lesson on CPU product development, and won't let Orochi mature as long as K10 did without coming up with a replacement.

Even if Orochi is still around to mature to the point where we are hitting 8ghz on air, it won't matter that current SB chips top out at 6Ghz. That's not the chip an 8Ghz orochi (if still around) would be competing against. Intel's aggressive Tick-Tock strategy will spew out something bigger and better by then.

Orochi dies need to perform now, and then be discontinued in favor of a new architecture if AMD is to succeed. The long term potential of each architecture if process maturity is allowed to proceed to infinity is really not important the way things are going in this market.

The only way it becomes relevant is if AMD/GlobalFoundries find a way to accelerate the process maturity in ways that have yet to be seen in semiconductor manufacturing, and that seems highly unlikely to me.
 
Zarathustra[H];1037807216 said:
Again, the world record doesn't mean anything.

If we exclude overclocking for a moment, in order for it to be a viable product it can't exceed a TDP of 125W. If AMD had a BD today that could clock higher than the planned FX 8150 and still not exceed 125W TDP then they would be releasing it, to make more of a challenge to Intel's lead. They don't, and that's why they have been forced to develop a smarter Turbocore system, but this is really more of a bandaid than anything else.

Now if we include overclocking, BD may be great for the few people in the community on phase change or chilled water solutions, but that's not very many of us. In the overclocking video AMD showed that their cherry picked BD sample was able to hit 4.7Ghz on a sealed water cooling solution. That's about the same clock speed as you get out of a Sandy Bridge with similar, but SB may have as much as a 50% IPC lead (if BD doesn't improve significantly on K10).

How well each architecture performs on extreme cooling like LN2 or Helium is utterly irrelevant today.

It may be important down the line once we see process maturity. As I recall Phenom II - when launched - would hit 4.2Ghz or thereabouts on cherry picked samples with extreme cooling when first launched. Now people are doing that on air, because of GlobalFoundries ridiculously mature 45nm process.

Hopefully - however - AMD has learned their lesson on CPU product development, and won't let Orochi mature as long as K10 did without coming up with a replacement.

Even if Orochi is still around to mature to the point where we are hitting 8ghz on air, it won't matter that current SB chips top out at 6Ghz. That's not the chip an 8Ghz orochi (if still around) would be competing against. Intel's aggressive Tick-Tock strategy will spew out something bigger and better by then.

Orochi dies need to perform now, and then be discontinued in favor of a new architecture if AMD is to succeed. The long term potential of each architecture if process maturity is allowed to proceed to infinity is really not important the way things are going in this market.

The only way it becomes relevant is if AMD/GlobalFoundries find a way to accelerate the process maturity in ways that have yet to be seen in semiconductor manufacturing, and that seems highly unlikely to me.
This is something I'm wondering, if switching from SOI to Bulk SI (HKMG) could help with some of the process issues that AMD had and that GlobalFoundries is having.

I have noticed one thing, that no matter if they use SOI or Bulk SI, no other foundry has the process ability of Intel, no matter how much they spend.
 
This is something I'm wondering, if switching from SOI to Bulk SI (HKMG) could help with some of the process issues that AMD had and that GlobalFoundries is having.

I have noticed one thing, that no matter if they use SOI or Bulk SI, no other foundry has the process ability of Intel, no matter how much they spend.

That I can't speak to. I'm a manufacturing/quality engineer but I am not specialized in semiconductor foundries, so I don't have detailed enough knowledge on it.

What it does show - however - is when you have large enough volumes (like Intel does) you can afford more engineers to work on the process, and the more engineers you have working on process design and continuous improvement, the better your process outcomes are going to be.

A pattern I have noticed as an engineer in manufacturing is that most manufacturing processes have endemic issues that inevitably lead to crises. When crisis occurs everyone rushes in to put out the fire, and work on process improvements gets delayed.

When you tackle the continuous improvements, you reduce the number of fires, which then gives you more time to work on more continuous improvements. So a lot of the time it doesn't take a lot more manpower to achieve a lot better results. Just a little bit to get you over the hump of getting out from under constant firefighting mode, and then the positive reinforcement cycle begins. This is where I am guessing Intel is today. Unencumbered by fires, and instead able to do the real work of improving their processes and thus yields.
 
This is something I'm wondering, if switching from SOI to Bulk SI (HKMG) could help with some of the process issues that AMD had and that GlobalFoundries is having.

I have noticed one thing, that no matter if they use SOI or Bulk SI, no other foundry has the process ability of Intel, no matter how much they spend.

Hmm you could say probably Intel is like Raegen was - forcing the speed of changes with which opponents cannot keep pace
 
Your entire argument falls apart when you learn that: BD has 2.4 Ghz reserve on SB and it's not even out yet. SB OC record is 6Ghz, BD's OC record is a world record @ 8.4Ghz.

Reserve ? If BD has so much untapped potential, why isn't AMD releasing a 4.5GHz stock SKU to at least beat the 2600k across the line ? Right now they seem to be stuck at 3.6GHz.

18 months ago I was saying that BD will not make it in 2011. It would be a really sad moment for AMD for that to happen ( in the sense of general availability, not press-launches ).
 
This is something I'm wondering, if switching from SOI to Bulk SI (HKMG) could help with some of the process issues that AMD had and that GlobalFoundries is having.

I have noticed one thing, that no matter if they use SOI or Bulk SI, no other foundry has the process ability of Intel, no matter how much they spend.

After doing some reading, I can't help but wonder if Bulk CMOS is a step backwards. It has more parasitic capacitance and less resistance to latch-up.

This ought to make it use more power at the same clock speeds, and thus get hotter than it needs to and not clock as high...
 
Back
Top