PCPER.com: 5 GHZ+ Phenom II Overclock on Dry Ice, 6Ghz on LN2

Printable View

Show 100 post(s) from this thread on one page

11-21-2008, 05:53 AM
Hans de Vries

Quote:

Originally Posted by Helmore

Uhm I'm not exactly getting what you are saying.
Are you saying that the NMOS driver current of 1354μA/μm on AMD's 45 nm. process is the same as on Intel's 45 nm. HKMG process, even though AMD's process does not use HKMG? Or is that NMOS driver current of AMD's 45 nm. process the same as what Intel would get if they had a 45 nm. process that did not use HK materials and metal gates?

I'm guessing the former....

Indeed.

Quote:

Originally Posted by Helmore

Could it be that their C3 revision of Shanghai and Deneb will implement these optimizations?

I don't know, but I guess not. I guess it would be a later revision.

Quote:

Originally Posted by Helmore

That could explain the lower TDPs on their AM3 Deneb chips.

Is it really lower? Once you need to increase voltage to go faster you
get a 3rd power law:

95W x (3.0 GHz / 2.8 GHz)^3 = 123W

So 95W for a 2.8GHz versus 125W for a 3.0GHz seems inline with this.

Regards, Hans
11-21-2008, 05:58 AM
Helmore

Quote:

Originally Posted by Hans de Vries

Indeed.

I don't know, but I guess not. I guess it would be a later revision.

Is it really lower? Once you need to increase voltage to go faster you
get a 3rd power law:

95W x (3.0 GHz / 2.8 GHz)^3 = 123W

So 95W for a 2.8GHz versus 125W for a 3.0GHz seems inline with this.

Regards, Hans

Ah ok.

I was talking about the fact that the AM2+ Deneb chip at 2,8 GHz. has a TDP rating of 125 W., while the AM3 version of pretty much the same chip has a TDP rating of 95 W. It's not exactly a night and day difference, but at least there is some improvement.
11-21-2008, 06:02 AM
informal

INQ(referrencing CRN) disclosed the exact max. frequency of 940 under LN2,it's 6.3Ghz.That's mere 11% lower than 7Ghz.I guess that extreme benchers will have fun time conquering the 7Ghz milestone now that 6.3Ghz was done on pre-production stepping.
11-21-2008, 06:02 AM
savantu

Quote:

Originally Posted by Hans de Vries

Thie abstract below was already a while on IEDM's technical program for
December. Not yet implemented on current Shanghai's but still to come.

Notice that a NMOS drive current of 1354μA/μm is the same as on Intel's
HKMG process without the HKMG. The current Shanghai has been
measured to have a PMOS drive current of 660 μA/μm. Much less as the
857μA/μm mentioned here.

http://www.his.com/~iedm/program/sessions/s27.html

Regards, Hans

AMD/IBM is 1354/857 NMOS/PMOS at what voltage and Ioff ?

Intel is 1360/1070 at 1V and 100nA/um Ioff.
11-21-2008, 06:25 AM
biohead

Quote:

Originally Posted by informal

INQ(referrencing CRN) disclosed the exact max. frequency of 940 under LN2,it's 6.3Ghz.That's mere 11% lower than 7Ghz.I guess that extreme benchers will have fun time conquering the 7Ghz milestone now that 6.3Ghz was done on pre-production stepping.

good lord this seems so unbelievable :shocked:
11-21-2008, 06:45 AM
Shocker003

Am not shocked by the news of AMD´s new cpu. They have been playing underdog for a long time and had all the time to rectify their mistakes. Well am waiting for these cpus to hit the shelfs. God 6.3Ghz for the underdog that has recieved a lot of bashing and the cpu was not cherry picked, got me burning with upgrading fever:bows:
11-21-2008, 06:46 AM
Stukov

Quote:

Originally Posted by informal

INQ(referrencing CRN) disclosed the exact max. frequency of 940 under LN2,it's 6.3Ghz.That's mere 11% lower than 7Ghz.I guess that extreme benchers will have fun time conquering the 7Ghz milestone now that 6.3Ghz was done on pre-production stepping.

Yeah, I was just gonna quote that if someone else hadn't. 6.3ghz, even if cherry picked is still pretty darn good. Much better than the 65nm Phenoms in pre-production.

I just wonder what the bad ole...er good ole P4 on 45 or 32nm would clock like.
11-21-2008, 06:52 AM
clonez

6.3 ghz....
damn, that is really fast, even for a post screen (which i expect, win stable would be too good to be true) :eek:
thats like double clock speed than the barcelonas @ launch:clap:

how long itll take until someone gets it bench stable @ 6ghz:cool:
i guess before 8.1 2009 :up:
11-21-2008, 06:52 AM
Hornet331

Quote:

Originally Posted by Stukov

I just wonder what the bad ole...er good ole P4 on 45 or 32nm would clock like.

yeah i wonder that sometimes meyself. :D

32nm HKMG p4... maybe 10 ghz? :p:
11-21-2008, 06:56 AM
clonez

more, given that the 65 nm parts managed over 8 ghz

i would place by bet at around 15ghz
11-21-2008, 06:56 AM
Donnie27

Quote:

Originally Posted by Sly Fox

Well there it is, I'm 100% convinced. :up:

Looks like just like with ATI vs Nvidia I was right... AMD is coming back!

And the best part? i7 is good too, options people!

Odd how when the news is positive for AMD SOOOO many "certain" users don't show up at all. What's up guys? Why no posty?

Here's why I didn't. As I told you, Hornett133 and even LOE;

Quote:

Originally Posted by BE

Damn , Charles no disrespect intended but you are sounding as cynical as Shintai & Donnie

Coming from anyone but him, I'd take that as a compliment. The other side of cynicism is loyalty and fanaticism.

I have a friend who's under NDA giving me what info he can and that's why I said I believed AMDin the other thread, remember? Doesn't change folks here with grudges uh?
11-21-2008, 06:58 AM
Hans de Vries

Quote:

Originally Posted by savantu

AMD/IBM is 1354/857 NMOS/PMOS at what voltage and Ioff ?

Well, It's not from IBM . It's a paper from the AMD Dresden guys who did
manage to do this. I presume they use the (more or less) standard
reference point of 1V/100nA/um. IBM tends to use 200nA/um

Quote:

Originally Posted by savantu

Intel is 1360/1070 at 1V and 100nA/um Ioff.

Yes, The very high PMOS current is because of their gate replacement
process. By etching the whole polysilicon gate away they leave a big gap
and all the strain is released on the channel.

Intel's 32nm HKMG numbers are: Intel is 1550/1210 at 1V and 100nA/um Ioff.
http://btbmarketing.com/iedm/release...news_final.doc

There are also some HKMG numbers from IBM/Freescale/AMD
1630μA/µm and 1190μA/µm have been demonstrated at 1V and 200nA/μm off
corrected for 100nA that would be 1570/1160μA/µm or so.
http://www.his.com/~iedm/program/sessions/s27.html

All still R&D papers of course at the moment.

Regards, Hans
11-21-2008, 07:02 AM
Particle

If I were the one at AMD doing a "partial handpick" I'd be testing about 20-30 random chips and make sure to pick something fairly average--not a lemon and not a cherry. Unrelated: What makes a lemon so bad and a cherry so good? I don't like cherries, but I'll eat a lemon anytime.
11-21-2008, 07:11 AM
D-Cyph3r

I saw this on Guru3D yesterday and couldn't get these forums to load at all.

All I could think was "holy :banana::banana::banana::banana: XS must have destroyed itself in the face of such epic news".

Anyways, birthdays in jan aswell. Deneb for meeeeeee!
11-21-2008, 07:15 AM
Ozzfest05

If they start making really high performance amd boards for this release I might pick one up on release although the intel boards are looking even better then before and they always have looked better then the amd offerings IMO, Ill wait and see
11-21-2008, 07:24 AM
Calmatory

Why would one need "better" board when the current ones can already hit these results?
11-21-2008, 07:27 AM
leoy

Top end 790fx costs 170 euros....
That gx tested only costs 120 eyros...:p::p::p:
11-21-2008, 07:27 AM
Particle

Yeah, I'd rather keep boards like we have now just with better PWMs. I don't need artwork, I just want a slab of PCB with some PCIe slots on it, some RAM slots, and a CPU socket. I care not what it looks like or how "simple" or "old" the layout is. Style can bite me. Give me an affordable board that can demonstrate good performance any day!
11-21-2008, 07:27 AM
Ozzfest05

Quote:

Originally Posted by Calmatory

Why would one need "better" board when the current ones can already hit these results?

IMO the current AM2 offerings arent very impressive im not saying they wont perform but I would rather go with the feature stacked intel 1366 boards which look nice offer performance and are stacked with features
11-21-2008, 07:30 AM
Hornet331

Quote:

Originally Posted by Calmatory

Why would one need "better" board when the current ones can already hit these results?

I think he means "better" board regarding to extras/components. (more sata ports, solid state caps, better power circuit etc... )
11-21-2008, 07:39 AM
DoubleZero

This will make life easier for the ones that like to support the underdog :)

Let's hope that AMD prices Deneb like Shangai, that is cheaper then Barcelona.
11-21-2008, 07:45 AM
Rammsteiner

Quote:

Originally Posted by STaRGaZeR

No, maybe you do know, not us. We are not blind believers, or at least me. I would never trust a hypothetical 8GHz Core2/i7 until I see the benches and the numbers. This is not a religion you know ;)

Im sorry to say, well, sort off, since that post is weird.

Why wouldnt a C2/i7 be able to hit 8Ghz? I mean, I never would have thought P4 couldt hit 8Ghz, but it did. Of course I know the architecture of a P4 is different then C2/i7, but still. I didnt expect in all fairness that i7 would go straight up to 4Ghz without a lot of issues.

Also a CPU-Z validation is worth just as much as a confirmation from multiple sources and people, like Tony and saaya, who should be most of the time unbiased in their expression.

Comparing Tony's post with a religion while your CPU-Z claim tells me that you do not believe it... Although there's been multiple times someone either managed to 'hack' CPU-Z or simply CPU-Z its self made an error. It's a nice program and most of the time it's right, still it ain't anything more reliable than multiple unbiased sources admitting the whole happening.
11-21-2008, 07:57 AM
Nedjo

who asked for pictures?

http://farm4.static.flickr.com/3064/...2863740558.jpg http://farm4.static.flickr.com/3014/...e16e6bc264.jpg http://farm4.static.flickr.com/3011/...961f6107f1.jpg
http://farm4.static.flickr.com/3210/...6e802f34e9.jpg http://farm4.static.flickr.com/3252/...6c52e9771a.jpg http://farm4.static.flickr.com/3040/...a5514a491d.jpg
http://farm4.static.flickr.com/3063/...f97e877777.jpg http://farm4.static.flickr.com/3295/...aaa5749a1d.jpg http://farm4.static.flickr.com/3290/...5f3b89cffb.jpg
http://farm4.static.flickr.com/3269/...6a493d027d.jpg http://farm4.static.flickr.com/3240/...b4fe0ce330.jpg http://farm4.static.flickr.com/3282/...e7d5da92ee.jpg
http://farm4.static.flickr.com/3062/...96d185c7ac.jpg http://farm4.static.flickr.com/3183/...808e43e771.jpg http://farm4.static.flickr.com/3024/...3da4befd62.jpg
http://farm4.static.flickr.com/3048/...21481f056a.jpg http://farm4.static.flickr.com/3008/...265a19869b.jpg

Macci rides again :D
11-21-2008, 08:01 AM
JumpingJack

Quote:

Originally Posted by Hans de Vries

Thie abstract below was already a while on IEDM's technical program for
December. Not yet implemented on current Shanghai's but still to come.

Notice that a NMOS drive current of 1354μA/μm (without HKMG) is the same
as on Intel's HKMG process . The current Shanghai has been measured to
have a PMOS drive current of 660 μA/μm. Much less as the 857μA/μm
mentioned here.

http://www.his.com/~iedm/program/sessions/s27.html

Regards, Hans

Hans,

THanks for the link... this is exactly what I would like to see -- unfortunately, I will need to wait for the proceedings to show up in the library some month or two later -- unless you go to this, in which case you can probably tell us what Ioff they extracted those number from (typically, that is 100 nA/um). The lower drive current in Shangahi is not really a concern, simply depends on the conditions. Typically AMD/IBM report AC Idsat numbers, and it would not surprise me if the chop shop measured it DC.
11-21-2008, 08:03 AM
JumpingJack

Quote:

Originally Posted by Helmore

Uhm I'm not exactly getting what you are saying.
Are you saying that the NMOS driver current of 1354μA/μm on AMD's 45 nm. process is the same as on Intel's 45 nm. HKMG process, even though AMD's process does not use HKMG? Or is that NMOS driver current of AMD's 45 nm. process the same as what Intel would get if they had a 45 nm. process that did not use HK materials and metal gates? I'm guessing the former....

Could it be that their C3 revision of Shanghai and Deneb will implement these optimizations? That could explain the lower TDPs on their AM3 Deneb chips.

That is what the abstracts says, and this is what Intel's 45 nm NMOS driver current is from last year's IEDM. It is hard to compare these numbers from this statement, the abstract does not specify voltage, or Ioff. If they did this at 1.4 V and an Ioff = 200 nA/um, for example, then this is siginficantly lower than Intel's work.

Show 100 post(s) from this thread on one page

All times are GMT -8. The time now is 01:04 AM.

XtremeSystems