Valencia and 16-core Interlagos are based on Bulldozer!

[MKruer]

How useful turbo boost is depends on a few factors that will be weeded out in time. The problem is that there is no good metric, nor is it a guaranteed performance increase. Likely yes, but you might get the runt of the Nehalem litter and the chip runs too hot or pulls too much voltage and the turbo mode never kicks in.

First, Is it sustainable? Can the chip run for extended periods time. I would argue probably not because if it could would you think they would market it as a fast chip?
Second, not all cores are equal, some of the cores might run a little hot or pull too much voltage there by shorting the "sprint mode".
Turbo mode both a gimmick and a good feature, but at the same time Intel is not guaranteeing that the turbo mode will be used, they are placing it in the gray area.

AMD solution is not any better or worse, its just different, and require every core to be able to sustain the same core speed. Sub sequential it requires the person to OC at the BIOS level, and uses the P states to down clock the chip speed.

BTW if you want to get technical the ability to OC is always software based by the virtue that it require programming instructions placed in the ROM to monitor and/or decide to allow OC's. A true hardware solution would require the absence of any such logic.

[keithlm]

"Sigh"
How about you quit with the intel spin....it's sickening.

Yeah... I had to look at the URL to make sure I wasn't at "The Hardware" forums. (Name obscured to protect the guilty. And noobish.)

POP QUIZ: If AMD implemented dynamic overclocking... and it allowed their chips to overclock a few hundred Mhz higher than the competition's highest speed for chips in the same price range... how many of the people that currently RABIDLY defend TurboBoost would suddenly change their opinion about using it on benchmarks?

And to be very clear: As I mentioned in another post: This feature is really ONLY about the benchmarks. Having a chip clocked at it's highest and then underclocking is a much better design.

[The_Ghost]

*sigh* just go work for AMD marketing already and stop pretending you are an objective engineering professional Abinstein.

You are full of it. Given the education you claim, you should know you are full of it. And you persist in attempting to mislead people away from objectivity and toward being AMD cheerleaders.

You are misusing your professional skills in order to shill for AMD and I dearly hope that someday I get to call you on it f2f in front of as many other members of our profession as possible.
(Shilling isn't the problem. Claiming to be objective while shilling =IS=. Just admit you're biased and not an objective professional and I'll have no problem.)

I'm done here.

you know, this is simular to what i use to tell you , go work for intel, because as far as i can remember, you have always acted like they was in your back pocket

maybe some day you will come up with enough courage to admit that you are also bias, the difference is that some people are not scared to admit it

where intel is killing amd, is with it's monopoly powers, the sooner that intel loses those powers, the sooner that we will see a fair fight

now that all this is said

please get back on topic, or stay out of this thread

i'm just hoping that i make this very clear

[duby229]

Lets take a slightly different point of view and ask what would it take for AMD to implement "Turbo Mode"? What mechanisms are missing?

I just happen to have a great answer for this question.... They can do what Intel does... Take a chip that passes the 3.2ghz bin, mark it as 2.8ghz on the box, but allow CnQ to clock it to 3.2ghz, then put an undersized heatsink on it so that it overheats at 3.2ghz and never actually gets there, and then sell it at a price premium........ Viola Turboboost....

[MU_Engineer]

Just as our body dynamically changes heart rate, BP, Blood sugar levels, etc to adapt to the "load" the "host" is presently under.

No external system can make such changes responsively enough or with enough sensitivity for mammals.
Same story for the degree and kind of control needed to properly implement dynamically changing CR for the transistors on a IC die.

Actually, guys, those physiologic parameters are very slow to adjust and most of them have a rather wide range of acceptable values. We can adjust them as quickly as the body can in many cases. Heart rate takes a little while to ramp up- several seconds in response to catecholamines to several minutes in response to increased muscular activity. Blood pressure is equally slow to adjust, but some mechanisms are very slow-acting to change blood pressure, such as the action of kidneys. Blood sugar levels are very slow-changing in healthy people as insulin release is very deliberate and human insulin isn't the most rapid-acting insulin available. Injecting a very short-acting synthetic insulin brings blood sugar levels down a whole ton quicker than anything the pancreas can do.

If you knew how to build a mammal, it would be less than trivial to build an external system to regulate their body functions.

It isn't exactly trivial to regulate body functions, but doctors are able to do so pretty decently for most things. We have at least the basics of many of the regulatory pathways figured out and can modify them artificially to keep somebody in the normal-enough range to keep many of them alive when the body has trouble doing it by itself:

- Blood pressure: fluids, diuretics, pressors, inotropes, alpha-1-adrenergic agonists and antagonists, phosphodiesterase inhibitors, etc.
- Heart rate: beta-1 blockers/agonists, antiarrhytmics, and electrical pacing.
- Glucose: various insulins, dietary modifications, exercise, glucose and other sugars given both orally and IV, glucagon, and a whole host of oral hypoglycemic/antihyperglycemic medications.

These are quite often able to be controlled on an outpatient basis by the patients themselves. Any diabetic that watches their diet, checks their glucose, and then gives themselves insulin is controlling their glucose using an external system.

[Lem]

Lets take a slightly different point of view and ask what would it take for AMD to implement "Turbo Mode"? What mechanisms are missing?

I just happen to have a great answer for this question.... They can do what Intel does... Take a chip that passes the 3.2ghz bin, mark it as 2.8ghz on the box, but allow CnQ to clock it to 3.2ghz, then put an undersized heatsink on it so that it overheats at 3.2ghz and never actually gets there, and then sell it at a price premium........ Viola Turboboost....

LOL, yes! Classic. Oh wait... Intel actually does that don't they, heh

[Smartidiot89]

Lets take a slightly different point of view and ask what would it take for AMD to implement "Turbo Mode"? What mechanisms are missing?

I just happen to have a great answer for this question.... They can do what Intel does... Take a chip that passes the 3.2ghz bin, mark it as 2.8ghz on the box, but allow CnQ to clock it to 3.2ghz, then put an undersized heatsink on it so that it overheats at 3.2ghz and never actually gets there, and then sell it at a price premium........ Viola Turboboost....

Did someone say Pentium 4?

[AzmountAryl]

Yes. I did. On May 11th 2008.

Nahalem will put Intel back in to the Pentium 4EE age

[Am486DX100]

"No external system can make such changes responsively enough or with enough sensitivity for mammals.
Same story for the degree and kind of control needed to properly implement dynamically changing CR for the transistors on a IC die."


I guess poison and physical/psychological trama mean nothing.

[rjon]

I guess poison and physical/psychological trama mean nothing.

Wouldn't physical trauma entail taking a sledge hammer to my computer!
I've never been to that point, but I know a few how have been close.

[Am486DX100]

"Wouldn't physical trauma entail taking a sledge hammer to my computer!
I've never been to that point, but I know a few how have been close. "




Videos of systems being thrown out of their owner's window.



*Sorry........Working my way back to the Topic.*

[skrubol]

If you knew how to build a mammal, it would be less than trivial to build an external system to regulate their body functions.

It isn't exactly trivial to regulate body functions, but doctors are able to do so pretty decently for most things.

You missed the first part of my sentence... If you had the [godly] ability to build a mammal, from the cell up, I'm pretty sure a task such as regulating its body functions would be trivial.

[scientia]

I see two different sides here. Kaa is correct that this is a factory endorsed and warranted overclocking. And, he is correct that overclocking less than all four cores is somewhat novel as it does give slightly more thermal headroom. However, I understand this as well from the origin and practicality points of view. I haven't seen anyone mention the origin but most here are probably arguing about practicality.

The orgin is something Intel would rather not talk about. Back in 2006 Intel produced a quad core chip at 2.66Ghz which was fine. However in 2007, they became afraid of competition from Barcelona and made an internal decision to push this to 3.0Ghz. But, as Anandtech's own data showed it was not possible to run a Kentsfield at 3.0Ghz with the stock HSF without exceeding the thermal limits. This naturally made Intel vulnerable to a class action lawsuit because their chips were not capable of running at the rated factory settings under normal conditions (as one would typically find in a consumer box). This was a direct violation of the Implied Warrant of Merchantibilty which all products have.

When the G0 stepping was released Intel's chips finally caught up the the factory rating and they began breathing easier. However they knew that this could happen again and some internal brainstorming began. Some here might know the origin of the "Baker's Dozen" which came about as the result of harsh penalties for baked goods that were below a certain weight. Some bakers might have tried to cheat their customers but most were geunuinely trying to hit the target weight. However because it was not an exact process it was possible that from time to time a customer could get slightly less weight. This would typically be balanced out though because it was just as likely that the customer could get slightly more weight from time to time. So, to get around the law bakers began stating a dozen but including 13 items. That way if the dozen were under weight the baker would say that the "free" 13th item put the total weight over the limit. In reality this process simply shifted the scale from a 12 count to a 13 count with the same likelihood of getting slightly more or slightly less weight for the total.

In many respects Intel's Turbo Mode is a Baker's Dozen. Intel can claim that the extra clocks are "free" and therefore shouldn't count if the processor is unable to reach Turbo speeds. In other words, if we turn this argument around we can see that Intel is selling a processor that is not guaranteed to reach its maximum clock speed. Note that this is exactly the opposite argument that Kaa is making about the overclock being factory warranted. Turbo is warranted but Turbo availability is not. An Intel fan at this point would typically say, "yes, but in most cases . . . ", and that shifts the discussion to practicality.

How useful is the ability to speed up one core? If three cores are idling most of the time then one would have to wonder why you would even need four cores in the first place. True, Turbo is painless overclocking. However, with AOD and the Turbo Button feature on Asus boards AMD's overclocking is nearly as painless. Frankly I think it is fair to bump a Phenom II by 200Mhz and compare this head to head with an Intel processor with Turbo enabled. Now, I understand that the extra 200Mhz is not warranted by AMD but I also believe that the risk is so miniscule that it is one that I would have no qualms about taking. I also understand that Turbo is warranted by Intel but this is balanced by the risk that Turbo will not be available because of core temperatures when using a stock case and stock HSF.

[Dresdenboy]

2. It has been said before and I will say it again. We will not have HyperThreadng in any of our products any time soon. There are places where it gives you better performance, places where it hurts performance. We will focus on more standardized places to deliver performance.

If that means AMD does not believe in any form of DSMT, then AMD is making a potentially company killing mistake.

Properly designed and implemented DSMT is the single best performance boost for CPUs in the last 15 years. And Intel is beginning to (finally) learn how to implement it right for x86 compatibles.

I warned AMD about P-M and the potential dangers of what became C2 and its descendents. I'm doing so again.

I invite you to my blog at http://citavia.blog.de/. I just updated it regarding SMT (linking to JF's comment in this thread BTW) and other forms of multi-threading. During the last days some patent applications have been published, which cover some multi-threading methods and how they could be implemented in an architecture having 2 integer clusters and a shared FPU.

However, Hans de Vries already stated (on aceshardware), that these new patent applications could also be related to a Bulldozer successor.

[JF-AMD]

I invite you to my blog at http://citavia.blog.de/. I just updated it regarding SMT (linking to JF's comment in this thread BTW) and other forms of multi-threading. During the last days some patent applications have been published, which cover some multi-threading methods and how they could be implemented in an architecture having 2 integer clusters and a shared FPU.

However, Hans de Vries already stated (on aceshardware), that these new patent applications could also be related to a Bulldozer successor.

I am not a AMD fellow. The fellowship program is a specific AMD program for engineers. I am on the business side of things, not the technical side.

If you want to write about this in your blog, I'd rather have us sit down and do a formal phone interview vs. taking comments of mine off the site. I think that would be more impactful for your readers as it would have much more of an "on the record" statement behind it.

We can't really comment about specific future products, but I would love to have a conversation about the future of server processors.

PM me and we can arrange a time for me to call you and we can do this formally. However, Ich spreche kleine deutsch, it has been about 25 years since high school and my german is really rusty

[Dresdenboy]

I am not a AMD fellow. The fellowship program is a specific AMD program for engineers. I am on the business side of things, not the technical side.

If you want to write about this in your blog, I'd rather have us sit down and do a formal phone interview vs. taking comments of mine off the site. I think that would be more impactful for your readers as it would have much more of an "on the record" statement behind it.

We can't really comment about specific future products, but I would love to have a conversation about the future of server processors.

PM me and we can arrange a time for me to call you and we can do this formally. However, Ich spreche kleine deutsch, it has been about 25 years since high school and my german is really rusty

I'd be happy to be able to talk with you about that topic. And we have a set of two languages to choose from, which should be enough More via PM.

I corrected my blog entry a bit right after reading your comment here. Unfortunately there was just not enough time to answer in this forum, but now I have internet.

Since my blog is about combining publicly available and sometimes less official information to discuss possible future µArchs of AMD, I just have to take into account plausible and sound hints which I find at different places.