AT / Gary Preview: 9850BE and new AOD

[SlowSpyder]

heyheybooboo, not sure if you know this or not, but I'll throw it out there. Based on what you said, it sounds like this will limit the Phenom more when trying to overclock the HT, as that affects all the system timings. So, do you know if, when using just the unlocked multiplier that is on the BE chips, does the new southbridge make a difference? I mean, I know it's best to push the HT/L3/IMC to get the most out of the Phenom, and it sounds like this will help you do that portion, but I don't know if it'll make a difference for your CPU's overall speed if you use the multiplier.

My chip makes it into Windows above 3GHz, but locks up the machine fairly quickly. At 2.9GHz stability isn't there at the temps I like. I doubt very much it's worth the hassle of changing my motherboard for another 2-300MHz, but my Phenom seems to be somewhat of a less then stellar overclocker (even by Phenom standards) anyway.

 

[lightstar]

i think this is really good news, hitting 3.4 GHz. . . . @ the $200 price point, the performance will be pretty close (or at least closer ). . . .if i were building a new machine today, i'd much rather support the underdog

 

[harpoon84]

Its definitely a step in the right direction for AMD, but am I the only one who sees this as 'too little too late'?

Its just that with Q6600s under $200 and capable of ~3.6GHz on air, its hard to get excited over a 3.2GHz Phenom unless you're an AMD fan. The reality is that we're still looking at a performance delta of ~25% after overclocking, which ain't peanuts.

 

[myocardia]

Originally posted by: Ocguy31
Explain to me what is so exciting about it and I will "shoo."

I will be excited when AMD puts out something that makes Intel say "Uh-oh", or "oh-sh*t", like they just did to NV. Id love to see Intel have to drop Nehalem prices due to a competitive market.

What's exciting about it to me is that AMD has no chance at achieving a performance parity with Intel's pre-Nehalem chips, unless they make improvements to both their CPU's and their chipsets. This may very well be the first step in an overall move toward that parity that they are so badly in need of having. Baby steps, man, baby steps.

I personally don't think AMD has to make Intel say "Oh sh*t", I think merely roughly equal performance to Intel's top-selling chips (not top-performing chips) would be enough to make Intel sit up and take notice. I don't think AMD will have any answer to Nehalem, unfortunately, so if you want a Nehalem, it's most likely going to cost you at least your first born child.

 

[myocardia]

Originally posted by: BLaber
Your questions were never misguided , but your comments after the questions you asked where not required on your part and thats what made me reply in the manner I did.

What's your point? This forum exists for three reasons: 1) to allow people to ask questions 2) to allow people to answer those questions 3) to allow individuals like us a place to share our opinions. And honestly, the opinions of a longtime, very well-respected member of this community, who also happens to have worked for a CPU manufacturer (not Intel, BTW) either designing or fabbing their CPU's, carries a hell of a lot more weight than the opinion of some noob, that has only posted in a total of two threads. No offense, I just thought you might like to be as cognizant of who you were conversing with as the rest of us happen to be.

Originally posted by: Idontcare
And here we are nearly a year after the first introductions of Barcelona and AMD is just now getting around to resolving the clockspeed problem imposed by these older SB's?

It seems wacky to me no matter how I look at it.

It seems a bit wacky to me, also. I'm sure the reason(s) are that either Hector thought everyone believed his lies, or they were deservedly spending that time getting their yield problems fixed on the CPU side of things, or most probably, some combination of those two. Of course, the reality may simply be that the problem that needed to be solved was more complex than they had originally thought, so they didn't devote as much money or manhours to it as they should have.

edit: Hah, that sentence was supposed to have read "...or they were deservedly spending that time getting their yield problems fixed, along with the TLB issue resolved, on the CPU side of things...". Thanks for pointing out my glaring omission there, Idontcare.

 

[Idontcare]

Originally posted by: myocardia
Of course, the reality may simply be that the problem that needed to be solved was more complex than they had originally thought, so they didn't devote as much money or manhours to it as they should have.

Having been there (not there as in AMD, but there in terms of debugging known problems on an under-resourced timeline) I can say this makes a lot of sense actually.

I hadn't considered the prospects of this having been an internally known contributor to clockspeed limitations but the pareto of such issues were such that this particular issue was ranked low enough that it's time for elimination hadn't come until now.

Actually this makes a hell of a lot of sense thinking back now, no doubt there was a significant allocation of resources directed to the top pareto item...the TLB issue.

Things make sense now, be they right or wrong I can at least accept the fact there is a perfectly plausible explaination for the outwardly laggard response to the southbridge limitations. Thanks for the help!

 

[aigomorla]

Originally posted by: myocardia
I'm sure the reason(s) are that either Hector thought everyone believed his lies

I just want to quote this.

You guys know my opinion on it so i wont say it because its gonna get me up for both parties.

But you know my opinion on it.

 

[Idontcare]

Originally posted by: aigomorla

Originally posted by: myocardia
I'm sure the reason(s) are that either Hector thought everyone believed his lies

I just want to quote this.

You guys know my opinion on it so i wont say it because its gonna get me up for both parties.

But you know my opinion on it.

Hector may have been one of those "leaders" who surrounds themselves with sycophants and thus he may have truly not realized they were lies when he was telling them.

Not that it is any less of an abandonment of duties by being that type of am executive leader.

The thing I just don't understand is that the market manipulation in Dec 2006 with that whole debacle over the analyst meeting that was done to secure whatever loans or funding they needed at the time only to turn around and have blow-out losses reported the next month...how did Ruiz and company avoid a sarbox issue with the SEC over that?

Or did the media play up something that really wasn't? There is a disconnect there, CEO's aren't generally allowed to lie to shareholders without some manner of punitive follow-up.

 

[ZootyGray]

An interesting speculative discussion people. But really, that's about why we are here waiting for the new findings from Anand /Gary. Don't get me wrong, I enjoyed the "ride" of your shared experiences. And it does set the stage for this hopefully successful adventure.

I think AMD is going to learn how to walk in this hostile environment. I have found it hard to accept that so much has just gone wrong. With the news re 750S/B, I have had the admittedly strange thought - very strange - that it has all been deliberate. That presents a whole new view; if you can go there - but, yeh, that's a strange pov. But look at the way NV was handed it's a$$. The usual disruptions of launch with NV raining on the parade was happening again - they were way over-extended. And the hammer dropped. I suppose I am talking super-consciousness in the marketing dept. Just thought I would share that 'strange' view. I am not only crazy; sometimes I am really out there. AMD has learned to shutup and do.

One quik question about the 'Idontcare' post - what means this = "HK+MG" ?? This discussion made sense to that point, then I tripped on the abbrev.

And it's nice to chat w/o all the fanboyisms and flames. I must admit I can and do go there; but I have my unique approach and I am passionate about my stand. However, I value genuine communication and mutual respect - not to mention an active intelligent discussion. I am not above gentle correction.

Thanks for sharing your experience. A very good read.

 

[ZootyGray]

Originally posted by: harpoon84
Its definitely a step in the right direction for AMD, but am I the only one who sees this as 'too little too late'?

Well, perhaps I am wrong; but having read the thread, I daresay you apparently are. And clock for clock doesn't really cut it anymore - it's an interesting comparison; but merely indicates something more distinguishes cpu's. For example, if core 4 is clocked higher than core 2 deliberately - then what - I surely do not know.

Additionally, this thread isn't about the cheap bargain chips that can be bought anywhere anytime. Thank you for that accurate information.

 

[heyheybooboo]

Originally posted by: SlowSpyder
heyheybooboo, not sure if you know this or not, but I'll throw it out there. Based on what you said, it sounds like this will limit the Phenom more when trying to overclock the HT, as that affects all the system timings. So, do you know if, when using just the unlocked multiplier that is on the BE chips, does the new southbridge make a difference? I mean, I know it's best to push the HT/L3/IMC to get the most out of the Phenom, and it sounds like this will help you do that portion, but I don't know if it'll make a difference for your CPU's overall speed if you use the multiplier.

My chip makes it into Windows above 3GHz, but locks up the machine fairly quickly. At 2.9GHz stability isn't there at the temps I like. I doubt very much it's worth the hassle of changing my motherboard for another 2-300MHz, but my Phenom seems to be somewhat of a less then stellar overclocker (even by Phenom standards) anyway.

It makes a big difference from how the Phenom has progressed over the last 6 months.

Gary got his Phenom 9850BE to 3.2GHz (200 x 16 x 12 x 10) - 3GHz at stock volts from what I can gather (or a 20-28% OC overall).

He doesn't detail the NB/IMC at 3GHz but if he is running 12x200 at stock volts that is a marked improvement. In my notes from the XS Phenom threads when upping the base clock (instead of using the cpu multiplier) the NB/IMC would have to be downclocked and the voltage jacked up. Wiith a 25% OC the NB/IMC was 8x (effective 10x) and 1.4v - and no amount of manipulation (or voltage) could raise it any higher.

Since the IMC is on-die any increase in voltage will generate more heat with which the cpu has to deal - further reducing headroom and stability.

The NB/IMC is more critical than the HT. It is the speed of the L3 cache and minor increases can effect performance quite dramatically.

Originally posted by: ZootyGray
~~snip~~

One quik question about the 'Idontcare' post - what means this = "HK+MG" ?? This discussion made sense to that point, then I tripped on the abbrev.
.

It is the transistor gate on a cpu. Previously all the 'gates' or switches were made of silicon dioxide. It's the next 'evolution' of transistor technology introduced by Intel at 45nm. It improves the energy efficiency of the transistor and reduces 'leakage' between switches.

 

[Idontcare]

Originally posted by: ZootyGray
With the news re 750S/B, I have had the admittedly strange thought - very strange - that it has all been deliberate.

Which part of it are you proposing was deliberate? There are a number of potential answers to this question and I'd prefer to not make any assumptions regarding where you are coming from.

Originally posted by: ZootyGray
One quik question about the 'Idontcare' post - what means this = "HK+MG" ?? This discussion made sense to that point, then I tripped on the abbrev.

As heyheybooboo wrote about, the term "HK+MG" is in reference to "high k-value gate oxide plus metal gate".

It is a process technology that Intel deployed with their 45nm node in which the high k-value (higher than the historically used silicon dioxide and more recently nitrogen doped silicon dioxide material) of the gate insulator enables Intel to reduce their xtor leakage (reduced power consumption) while maintaining relatively higher clockspeeds.

The high k-value dielectric is not without its drawbacks, notably it decreases the mobility thru the channel which has the undesired effect of reducing your chip's clockspeed capability. To counter-act this deliterous mobility issue one replaces the gate material (historically doped polysilicon) with a metal gate (in reality a stacked or layered structure of metal-based materials such as TaN, etc). The metal gate improves the mobility in the channel, enough to counter-act the mobility degradation caused by the integration of high k-value dielectric.

Not that adding metal gates is without its own set of drawbacks.

You will also see the technology refered to as "hi-k/MG" as well as "HK/MG".

AMD has publicly stated their initial products on 45nm will not use HK+MG, rather they will stick with the traditional transistor materials of nitrogen doped silicon dioxide for the gate oxide and doped polysilicon for the gates.

 

[Shortass]

Originally posted by: Ocguy31

Originally posted by: Shortass

Originally posted by: Ocguy31
Woohoo! 3.2ghz 65nm Quad-core! Lets party like its Jan, 2007!

I am genuinely excited about this development, and your sarcasm is hardly appreciated. Shoo, troll, shoo!

Explain to me what is so exciting about it and I will "shoo."

I will be excited when AMD puts out something that makes Intel say "Uh-oh", or "oh-sh*t", like they just did to NV. Id love to see Intel have to drop Nehalem prices due to a competitive market.

As far as a 65nm quad-core than can maybe hit 3.2ghz on the right board....that rates right up there with a dentist visit on the excitement scale.

It only rates that low on the excitement scale if you lack the ability to think farther than your arm's reach. With AMD showing that they are fixing various platform issues and are ramping up clock speeds finally it isn't hard to fathom the die-shrunk variant hitting even more impressive clocks. With any amount of further chipset advances and optimizations they might become very competitive with Intel in the mid range by their next gen of cpu's.

If they're even remotely competitive by that time, Nehalem will like drop in price MUCH quicker to encourage sales and limit AMD's cash flow. Even you should be excited, you'll be able to retain your pathetic fanboy status at a cheaper cost!

 

[OCGuy]

Originally posted by: Shortass

Originally posted by: Ocguy31

Originally posted by: Shortass

Originally posted by: Ocguy31
Woohoo! 3.2ghz 65nm Quad-core! Lets party like its Jan, 2007!

I am genuinely excited about this development, and your sarcasm is hardly appreciated. Shoo, troll, shoo!

Explain to me what is so exciting about it and I will "shoo."

I will be excited when AMD puts out something that makes Intel say "Uh-oh", or "oh-sh*t", like they just did to NV. Id love to see Intel have to drop Nehalem prices due to a competitive market.

As far as a 65nm quad-core than can maybe hit 3.2ghz on the right board....that rates right up there with a dentist visit on the excitement scale.

It only rates that low on the excitement scale if you lack the ability to think farther than your arm's reach. With AMD showing that they are fixing various platform issues and are ramping up clock speeds finally it isn't hard to fathom the die-shrunk variant hitting even more impressive clocks. With any amount of further chipset advances and optimizations they might become very competitive with Intel in the mid range by their next gen of cpu's.

If they're even remotely competitive by that time, Nehalem will like drop in price MUCH quicker to encourage sales and limit AMD's cash flow. Even you should be excited, you'll be able to retain your pathetic fanboy status at a cheaper cost!

Ah...a personal attack against someone who is pointing out facts. Not going to score any points in a conversation like that.

Your whole point is based on assumptions of possible future improvements by AMD. I am merely commenting on the current situation, on a chip they are releaseing now, which is the topic of this thread.

 

[ZootyGray]

@Ocguy31
(yawn) The only attack is yours. Shortass has succinctly offered you clarification you requested - he has not mentioned 'victim mentality', 'paranoia', 'self-defeating adolescent chest beating', 'other related ego-based delusions' (or limitations related thereto), inferiority masquerading as superiority', or 'miscellany of neuroses' (common to incendiary indulgences generally), or the fact that these (or analagous) are 'the essentials' for such identification. However, your position is clear; thank you for that.

 

[ZootyGray]

re HK+MG
Thank you for response to my little quandary. I do recall reading up on that tek; unfortunately, I did not finish my reading at the time, wrote it off as "later" - and here we are.

@ Idontcare - an excellent, and efficaceous, description mid-thread. Thank you and pardon the interruption; I daresay perhaps others have also been quickly updated. It's on me to read more on that.

@ Idontcare - Also - where I am coming from re "deliberate". That's hard to explain in great detail. I mentioned that I can really get way out there sometimes. Trying to explain creative or revelatory vision or such glimpses or flashes in free-thought drifting can be highly misunderstood, or even dangerous, subject to ridicule. This thread had already approached the "deliberate" notion (it seemed to me), and that triggered this earlier thought I had had.
=I recall the thought moving into the idea of buying valuable stock for seriously cheap. =The illusion of this underdog seizing power when everyone thought it was going under is a vision of triumph. =And, as I mentioned before - the hammerdrop on NV - a similar notion to let ntel overextend and pull the rug out with Deneb (or what?) or perhaps this report we are awaiting from Anand/Gary. =The underdog idea and the hope that AMD will survive and become the best. =The fact that AMD seems to always pull more trix out of nowhere.
Really, this my personal fantasy. It is related to fact; but is largely an unbridled extrapolation. It makes no sense to me that a company with great assets should just go fizzle and pop. I expect to see gpu/cpu/multicore merge (the all-in-wonder-mobo) and AMD is the only one that can do it, at this point in time. ntel doesn't really have vid/grafx. AMD claims true quad (I don't have the facts on that either btw). (disclaimer - This is my personal ramblings - I could go on and on - but I am not stating fact necessarily; just thoughts I have had).

Much of this is just a hopeful AMDfan cheering on the little engine that could I have stated elsewhere that I am a self-confessed AMD fanboy for ethical and personal reasons. I want AMD to do well because I am shopping for a new pc right now - and I want problems out of the way. My new pc will be an AMD-based unit that I built. I have designed and repaired pc's since 486 era. I am self-taught (stole it from everybody) and I read a lot of stuff that most people just won't. Anyway - too much about me here.
(Behind it all - it says 'don't you dare underestimate what we can do. We will deliberately go here, do this, cos we have to; and then we move when prepared.') That's an old concept - a warrior waits, and a warrior knows what he waits for. (humility)

Dreams. I am not just crazy - it's much worse/better than that.

This is a good thread - I apologize for potential side-track /hijack. I only want to respond to the question posed to me. (continue)

 

[Idontcare]

Originally posted by: ZootyGray
largely an unbridled extrapolation

I hear you there brother, the concept of hope is something we homosapiens have that others in the animal kingdom don't.

I too hope AMD pulls some proverbial rabbit out of a hat here, not for the sake of the company or its shareholders but for the sake of the thousands of employees who have families depending on them to provide some degree of financial stability and sanity.

 

[ZootyGray]

I find it very exciting that AMD, in midstream, (!) is almost redesigning the function of Phenom, with the new bios, new oclock utility, and this new "low-level" ACC.

Since the early days of my study of pc's, I always wanted to know what makes these things work. Today I can say I am naturally drawn to things like "low-level" or "machine language". The individual (hello Tom Terrific!) who, in my early days, used to sit me down beside him at his pc, and say "watch and learn", told me that there were only a few people in the world who could write machine code. While we will not get to do that, the idea of tweaking the inner workings of hardware through a possibly powerful software interface, really has my attention. And that AMD might be making new areas accessible is hugely exciting - I am probably going too far here - but that's where my interest is.

This is like being in some waiting room; wondering and waiting, and passing time; waiting for the announcement. And the statement that they are talking with AMD brings it all closer. How would you like to be a fly on the wall for that chat?

 

[formulav8]

My lord, I am really sick of these forums lately with nothing but people who act like Intel drones. They know who they are... :disgust:

Intel DOES NOT like any of you. Get a life people. Its only a stinking computer.


Jason