HKMG overated?

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

yomamafor1

Distinguished
Jun 17, 2007
2,462
1
19,790


So let's see, they've gone from sub-par 65nm operation to on-par 65nm operation. Yeh that's an improvement, but simply not enough.

If you're so kin on C1 or C2 to achieve magical power consumption, show me a CPU-Z shot. If you're so sure that Deneb is going to be on par with Intel's 45nm technology, show me a CPU-Z shot.

You know, when a person deduce a conclusion from an observation, he/she's called scientist; when a person deduce a conclusion before he/she even sees one solid fact, he/she's called fanboys.
 
Im sorry, did I not mention that things should be kept in context? Reread the post. Ill ask again, show me where it wont work, cause IBM and AMD and TSMC and et al need to lnow this.

Is this a theory or fact that it wont work well. Didnt mention that part. I could research more on peoples daft comments here about HKMG, but I simply wont, as everyone else knows how it really is. Im just here to point a few things out is all. If I were you, I wouldnt hang my hopes on HKMG as making something work well vs work better. Ones negative (Intel camp) and ones positive (everybody else) And why was it negatively portrayed? Hype? I was told anything at 45nm would leak like a sieve, but thats not true is it? On an older revision of a chip thats not released yet, it gets as good as a 65nm Intel chip, and supposedly by all the HKMG belevers out there it wont work as well as the preious gen, ya know, Phenom?
 

NMDante

Distinguished
Oct 5, 2002
1,588
0
19,780


NO, you are adding HKMG into the statement. That is YOUR addition, not anything anyone has claim, as your topic and what you have been arguing about. NO ONE has said HKMG is needed for anything. You are the only one who is claiming that. YOU. No one else. You are simply putting your own thoughts in other people's comments, and claiming they said that. They didn't.

How exactly does "45nm SOI simply won't work well for them" mean "AMD needs HKMG"? It was an opinion that SOI and 45nm won't work well. I don't see where HKMG comes into play, yet you add it in there. Based on...? Nothing. In fact, the comment was a reply to this:
I don't think 45nm will do anything for them.

First generation will use SOI. They are still working out the bugs and problems with K10.

This just has me doing what I was thinking about doing all along... X48+DDR3+Q9450 in the Summer.. If AMD can't compete Intel isn't going to bother.
So, that is all about HKMG, too?
 

yomamafor1

Distinguished
Jun 17, 2007
2,462
1
19,790


Its called resource allocation.

p.jpg
 
And about TSMC? And IBM? All have been mentioned along with HKMG. It tells me, Intel fans are overhyping something that doesnt show great impact at 45nm, let alone 40nm. It showws me alot of people bought into something thats a help, and not a need, as witnessed by all these comments. But, as theyre mentioning it, its not as a help, its dire that its done, or else. Cant find the word need, nor did I bother to look. But the implications are there.

All this came at the expense of AMD,IBM and TSMC. If someone does as good without it, will you be convinced? If so, hang on, and we will see. If AMD, and what that earlier slide shows, proves its already equal without HKMG. Will it be better with it? No doubt, but again, whats that telling you?
 

yomamafor1

Distinguished
Jun 17, 2007
2,462
1
19,790


Not work well != not work.

p.jpg

 

NMDante

Distinguished
Oct 5, 2002
1,588
0
19,780
Geez, you love to use "hype" a lot.
Simple - IBM made statements about using HKMG, the exact same time as Intel. So, was IBM "hyping" HKMG? According to you, they must have been.

Again, if IBM or TSMC thought HKMG has no use, why are they still researching to use it?

Also, you are now basing AMD's 45nm performance and power consumption on slides? Okay. How many 45nm AMD CPUs have been reviewed? Go look at the Shanghai review at Anandtech, and see if there was huge power efficiency gains and performance.

You are speculating on whether or not AMD will be better off without HKMG, with no data. What is that tell you?
 

yomamafor1

Distinguished
Jun 17, 2007
2,462
1
19,790

Doesn't show a great impact?

4W idle for quad core not a great impact?
2.5W idle for dual core not a great impact?
22.7W load for dual core not a great impact?

Then what, in your opinion, is a great impact? Running 4.0Ghz with 1.6V? Overclocking to 4.0Ghz according to a slide?



Again, proof?

AMD also showed 3.0Ghz 65nm Phenom to the press. We all know how well that worked :sarcastic:
 
Am I? Or, am I just pointing to the usefulness or non usefulness of HKMG? You say it wont work well. According to? Links? Proof? Ive said time and again, itll work better, but theres a world of difference in how Im saying it, and how youre saying it. If Im so inclined, I may look up a few comments on it, and see just whats been said, as I know how it went, as do others. And on it still goes.

If its proven by AMD that the same clocks can be achieved without HKMG, then that disproves alot of bull thats been spread. And its looking like I may see it happen.

If this does happen, Im sure many people wont want to hear about this, and others would just as soon forget about all the bull that came before it. Im just here, at this particular time, to remind everyone about this, and its implications.
I think HKMG is a very fine process, and will help with leakage at lower nodes. But, you wont be finding me running around saying this manu should have it or it wont work well, because, simply, this just may not be so. Period. Work better? More than likely, but not to work well? By whos standards? And if its as good as Intels standards without it? And a Phenom die shrink, first time at 45nm? Lets see whats going to happen, and not just AMD either. Lets wait for when TSMC surpasses Intel at the 40nm node, and see then. Im not that familiar with IBM, so I cant actually say there, but let us see what happens, because if the lowly Phenom, which cant be oceed, and came in at 2.6 tops, goes to over 4Ghz on air without HKMG, then Id have to say, it wont work well is misinformed, not thought out,bashing
 

yomamafor1

Distinguished
Jun 17, 2007
2,462
1
19,790



If you cannot pull your head out of the gutter, I can NOT help you.

p.jpg


EDIT: Let's break it down real quick

What's proven?
- HKMG helps with leakage
- Intel's 45nm CPU has very low power consumption, both at idle and load
- 45nm SOI requires high voltage to clock high

What's not proven?
- AMD's 4.0Ghz overclock
- AMD's C1/C2 stepping

'Nuff said.
 
Personal attacks just dont cut it with me. So get that right now.

So, what happens when you have massive leakage? Whats the end results? Heat? Power usage? And... no ocing, especially on a core thats shown bad ocing in the past, barring the shrink. So, my points made if it ocees to those numbers, and soi running at high speeds without HKMG wont be a theory, will be working as well as Intels with HKMG. Like weve all been told, it cant ocee that well with soi and without HKMG, time will tell
 

yomamafor1

Distinguished
Jun 17, 2007
2,462
1
19,790
Selectively ignoring factual information also don't cut it with me. So get that right now.

Have you noticed something? It seems that most of your argument is based on "if".
 

NMDante

Distinguished
Oct 5, 2002
1,588
0
19,780

Sorry, but you are using what I said as fact?
I simply said that it was theorized that HKMG would be used at 45nm, because IBM said it was planning on using it at 45mn. Now, I didn't realize that what I say would be fact, and not MY opinion.
Thanks for granting my opinion as fact, but it's just an opinion. If you can find any other claim, other than my own opinion, claiming it was a theory, then that's fine and dandy. But again, you are taking my opinion out of context.
It was mentioned that HKMG was going to be needed after 65nm, since power leakage would be substantial (theorized) at 45nm and lower, without it. That was a theory. IBM was going to use low-k MG at 45nm and go to high-k later. Again, if you believe high-k to be overrated, why would IBM consider using it, if it was not needed at all?
Now, where did I claiming this was fact? Did I link any articles to my opinion? Nope. It was a theory. My theory. There was leakage for Intel at 65nm, and they looked at different methods to fix it. One was SOI, the other was HKMG. Again, a theory. My opinion, based on what I had read about HKMG and it's advantages for transistor performance and power consumption, from IBM and Intel press releases.

Unless you have some other article claiming the same theory as I had, then you have no proof of "fact to theory". You have proof that I believed it was a theory, of why HKMG was needed going from 65nm to 45nm.
 
So, oncee again here : "The transistor drive current for AMD's 45-nm devices is much lower than that of the Intel HKMG transistors. But power consumption is quickly becoming a high priority for server chips. AMD's transistors exhibit very low channel leakage. Our transistor benchmarks indicates that leakage current is less than one-third of the value measured on AMD's 65-nm process. It's also significantly lower than the Intel 45-nm HKMG process. In fact the Ion/Ioff ratio for AMD's PFET is nearly 10 times better than that for the Intel PFET. "
No ifs and or buts. No theories, no not work well, none of that.
 

Claimintru

Distinguished
Aug 4, 2008
68
0
18,630
Jaydeejohn what are you even arguing anymore? SOI is outdated technology, in sub 45nm manufacturing it won't exist.

AMD said it won't compete in the high end market, because it can't. Intel doesn't make a majority of its profit from the high end market, what they do make there however are people's recommendations. A year after the Core 2 Duo processors came out when I was working retail people would come in who knew absolutely nothing about computers, asking for Intel over AMD just because what they heard from "their friend who knows these things"

What company is in the red what company is in the black?

The ATI 4xxx series = amazing having low price and high performance, but it still wasn't enough to get AMD back on track, no where close. Why even bother arguing about any of this anymore? Scroll up, you're ranting about HKMG gates and how you'd love to mouth AMD's and IBM's shafts collectively, and have been for the past 10 posts.
 

NMDante

Distinguished
Oct 5, 2002
1,588
0
19,780
What are you talking about?
I based my theory about HKMG @ 45nm based on press release information from IBM and Intel.

You are basing your "ifs" on unconfirmed rumors and marketing slides.

You say, what if AMD's 45nm can overclock to 4Ghz, then HKMG is not needed. Okay. But if AMD uses HKMG in the next generation of 45nm, what does that mean? That it was needed?

If you really want to prove your point, find the post/reply/thread that clearly states that "AMD needed HKMG @ 45nm", which is the basis of your entire argument.
 
No, what Im saying is, you along with others have claimed you cant get high clocks without HKMG. You along with others have said things like itll be a burner without it, and will never oc. You along with others have said the thermals would make it impossible. Im saying, if it happens, you and all the others are wrong, were wrong from the get go, HKMG isnt a must have, and all the crap made up about how anything without it will not work well is wrong. I dont care if VIA had it first, or AMD or whoever, and whever Cometoo is, is wrong as well . If its used, it helps, but all the theories about soi not working well without it at 45nm is going to be proven wrong