Phenom 9500 w/ MSI K9A2 Platinum

[aGeoM]

Guys I sold the K9A2 with Team Group 4Gb memory modules @1000 5-5-5-15 2.0v and Phenom 9600 @2500 HTT/NB x9, 8h Prime stable. They are very good for 70 euros.

[batmang]

I'm going to try XP64 and see how much farther I can get.

[tictac]

If you had difficulty downloading collection of the MSI K9A2 Platinum beta bios.

I got the beta bios collection on one page over here: MSI K9A2 Platinum Beta Bios Collection

Happy Tweaking and Overclocking

[KTE]

dr_drache Yeah I should be able to check 'em out, let me first get hold of a Phenom first. No B3 here yet.

Kayin: Geil, what part number is that? I can't get hold of that here that I know, no where.

aGeoM: Thanks for the heads up

batmang: Should be better IME than Vista 64b anyway.

Cutie: Great collection you have there, thank you mate

[aGeoM]

If you had difficulty downloading collection of the MSI K9A2 Platinum beta bios.

I got the beta bios collection on one page over here: MSI K9A2 Platinum Beta Bios Collection

Happy Tweaking and Overclocking

2 more:
A7376ACI V1.0B3/A7376AMS VP.0BG - http://www.xtremesystems.org/forums/...postcount=1524

dr_drache Yeah I should be able to check 'em out, let me first get hold of a Phenom first. No B3 here yet.

Kayin: Geil, what part number is that? I can't get hold of that here that I know, no where.

aGeoM: Thanks for the heads up

batmang: Should be better IME than Vista 64b anyway.

Cutie: Great collection you have there, thank you mate

Welcome, BTW the system was with XP x64, and bios A7376ACI.103, I could not set Promise RAID, neither CD integrated or by F6 FDD, had to be in SB600.

Hope to see some B3 results, B3 in Germany by April 2.

[KTE]

Yeah some parts of EU and US get them far before us each time. MFGs usually play games by doing paper launches in most of the world but satisfying just one market; mostly the US market.

[batmang]

Well... XP is definitely alot more stable than Vista when overclocking... thats for SURE. I'm at the same overclock I was in Vista64, but I'm at a lower voltage. 2.6GHz is still unstable, but I was able to get 13.5 (2.7GHz) multiplier for a few seconds then it froze... :P 2.5GHz is fine to me, heres the latest. I'll rerun 3dmark06 later today.

http://www.jmbat.com/media/phenomoc/...64_2point5.JPG

I'm running XP 64 btw.

EDIT: Bumped the vcore to 1.26, got a random lock up. :\
http://www.jmbat.com/media/phenomoc/...int5_11197.JPG

[madfaze]

just noticed when you adjust the NB volt it bumps its PCIe volt... :

e.g.

NB volt = 1.2v
PCIe = 1.325

but when NB volt is in stock reading of PCIe is only 1.250v...is this safe?anyone can confirm why it does?

used dual core center to read it..

[SocketMan]

I'm going to try XP64 and see how much farther I can get.

http://www.xtremesystems.org/forums/...d.php?t=179058

[batmang]

http://www.xtremesystems.org/forums/...d.php?t=179058

Makes sense NOW. My system became unstable at 2.5Ghz randomly. I could play games for a while, randomly would reboot doing not so taxing things. It's back to Vista 64 and no overclocking for now. I'm not going back to XP 32bit, no way.

[KTE]

I've seriously had information overload for the past 5 weeks. So much info, no time to menton or explain.

Wanted to post some BIG pics of nearly all new AM2+ MBs, check 'em out, 3-way SLI boards here too! To list just a few:

AMD 790FX
Albatron KX 790FX

AMD 790GX
DFI LanParty DK 790X-M2R

AMD 780G
Abit AN78
ASUSTeK M3A-H HDMI
ASUSTeK M3A78-EMH HDMI
Biostar TA780G M2+
ECS A780GM-M
ECS A780GM-A
Foxconn A7GM-S
GIGABYTE GA-MA78GM-S2H
Innovision A78GM-HDMI
Jetway BA-200-KU
Jetway HA06
J&W Technology JW-RS780GA-UVD
MSI K9A2GM V2

AMD 780V
MSI K9A2GM-F V2

AMD 770
ECS 770M-A
Foxconn A78AX-S
Jetway HA03-GT
Jetway PA77GTA-VT
Jetway BA-10

AMD 740G
Biostar A740G M2+ SE
ECS A740GM-M
Foxconn A74MX-S
J&W Technology JW-RS740G-DVI

Nvidia nForce 780a
ASRock K10N780SLIX3-WiFi
ASUSTeK CrossHair II Formula
ASUSTeK M3N-HT Deluxe/Mempipe
FOXCONN DESTROYER
GIGABYTE GA-M780SLI-DS5
MSI K9N2 Diamond

Nvidia nForce 750a
ASRock K10N750SLI-WiFi
ASRock K10N750SLI-110dB
BIOSTAR TPower N750
J&W Technology JW-NF750A-HYBRID

Nvidia nForce 730a
ASUSTeK M3N78-EMH
ASUSTeK M3N78-EH Deluxe 730a
J&W Technology JW-NF730A-HYBRID - nForce 730a

Nvidia GeForce 8300
ASUSTeK M3N-H/HDMI
Galaxy Technology GeForce 8300/nForce 730a
J&W Technology JW-G83UM-PVHD+

Nvidia GeForce 8200
Abit AN78G
Abit A-N78HD
Abit A-N78H
ASRock K10N78hSLI-WiFi
ASRock K10N78hSLI-1394
ASRock K10N78FullHD-hSLI
BIOSTAR TF8200 A2+
BIOSTAR GF8200 M2+
ECS GF8200A
ECS GF8200M-M3
FOXCONN M78A-S
Galaxy Technology GeForce 8200/nForce 730a
J&W Technology JW-G82UM-PVHD+
Jetway HA05
Jetway PN78GTA-GHG
Jetway PN78GT3-GHG
MSI K9N2GM
MSI K9N2GM V2

Alot of them are here: http://digitallife.jp.msn.com/articl...icleid=273449/

You can find many of them here that aren't above, close up, far better than the above: http://bbs.danawa.com/view.php?nSeq=...logCateSeq1=32

I've uploaded some bigger, closer-up pics on imageshack, check 'em out according to the tags: http://profile.imageshack.us/user/kte/

Review of ECS 780G vs MSI 690G vs Palit 8200:http://www.digit-life.com/articles3/...gm-a-780g.html

Palit 8200 vs Biostar 8050: http://www.digit-life.com/articles3/...force8200.html

Interview with AMD: http://www.tomshardware.co.uk/AMD-Ph...iew-30449.html

Tom’s Hardware: When only raw performance is considered, Intel is currently in the lead with its desktop processors. The competition has also made great strides where energy efficiency is concerned. How do you plan to retake the lead in this area?

Polster: Well, the launch of AMD’s Phenom using the B2 stepping did not go quite as we had imagined. This will change with the B3 version, which will be used both in our quad-core and the triple-core CPUs. But let’s look at the facts for a moment. The Phenom is selling very well in the retail/end customer market, and we have absolutely no reason to complain. However, the OEMs and system vendors are causing us a bit of a headache. This segment is waiting for us to release the B3 stepping.

Tom’s Hardware: Is there any truth to the rumour that the Phenom contains a bug in its third core?

Polster: None whatsoever.

Tom’s Hardware: Our tests have shown that the Phenom is not as efficient in idle mode as it should be. Despite the improved Cool’n’Quiet 2 power saving functionality, the CPU’s idle power consumption is not exactly impressive. Here are a few idle power measurements we determined in our lab: AMD Phenom 9600 Black Edition – 20.27 Watts, Athlon 64 X2 3800+ - 8.42 Watts, and Intel Core 2 Extreme QX9650 – 3.78 Watts. How do you explain the Phenom’s comparatively bad results?

Polster: I’d say we’d have to take a closer look. Power consumption should be much lower with Cool’n’Quiet 2 enabled.

For reference, the AMD RD790 chipset bundle is a 11.4W TDP max and 8W max. average with 1.xW TDP idling chipset (incl. Southbridge). The GeForce 8200 chip has a 14.3W TDP alone.
Anyway, check what AMD officially said of CnQ 2.0 power consumption for Phenom.

Unfortunately the intel linpack binaries do not work on the phenom system, so i could not yet try it, must build my own binaries first and i'm very short in spare time atm.

I tried that and it didn't work. Got a developer to build it for me too and it still failed to run (wrong platform).

I used an qx6850 GO ES did not play with the FSB, due to lack of time. With an 11x multi and 333MHz FSB wattage is ~230W DC during prime95 test.

Whats the stock VID and Vcore on it?

Saw that news. Those xeon's are for dual socket systems, and they still use fbdimm's, that might level things out abit. Does 7-year life cycle mean 7-years warranty?

Yeah, that's one major problem, their chipset and their platform doesn't match in terms of cost/energy efficiency. Their 7300 chipset for Tigerton I played with at work today for a bit, it has 4 separate FSBs (DHSI) for each socket and 8 DIMMs for each socket too.

Have you checked this out? http://aceshardware.freeforums.org/c...iler-t428.html
You can skip most of it as fanboy drivel but some parts are very relevant to coding; Intel (in its best compiler) is using a separate path for AMD CPUs than it does for its own CPUs while that separateness is not needed if both were optimized the same. So the optimizations for Intel CPUs actually run code much faster than those for AMD CPUs.

madfaze: Which BIOS did you change volts in? What does AOD read?

Kayin: Esoteria CAS3 800 didn't fit side by side in dual channel, they were too thick. There was also no supply of them here, I tested on someone else's system. The EVO's are available here though...

BTW how many of you realize that we've already covered individual core clocking without AOD about 5 weks ago?

[KTE]

batmang: 64b is a known problem for many.. I haven't worked out why and don't have any answers to it either.

You can find many AM2+ board pics/specs here too: http://products.amd.com/en-us/MotherboardResult.aspx

Here's two AMD 780G videos I forgot to post:
http://www.youtube.com/watch?v=Kd0Of4PnpQk
http://www.youtube.com/watch?v=V741r...eature=related

Leslie speaks about the new AMD lineups (not that you can see her ):
http://www.youtube.com/watch?v=ZQ_MV81yoMc

"AMD is planning to launch B3 stepping quad-core Phenom 9050 series CPUs including 9850, 9750, 9650, 9550 and 9150 within the next 1-2 weeks, however the top-end 9850 will only reach a core frequency of 2.5GHz, instead of the original goal of 2.6Ghz [....] AMD will later launch triple-core Phenom 8750, 8650 and 8450 CPUs with power consumption of 65W in late April. Prior to that, the company will launch the Athlon 64 X2 5800+ to fill the gap in the mainstream market, noted the sources".

I didn't know that, if true, then you might have a big OEM/vendor hit

[JumpingJack]

Now that shows nothing really. What can you expect from ITSA 45nm?
First of all, far lower wiring delay and power leakage due to Ultralow-k use, more than that of Intel 45nm, and secondly, IBM/AMD already presented their experimental findings of the highest PMOS (110) transistor performance with only using conventional SiON + compressive liner and eSiGe stressors with optimized Rext. at Lgate=35nm, Vdd=1.0V, 250 nm poly-pitch: Ion over 1 mA/μm at Ioff 100 nA/μm. AFAIK the concentration of pFET Germanium was <30&#37; and SiC was used for nFET performance advancements. This was back in September 2007, so quite obviously with HK/MG, even with metal gate-first approach, they are not going to be much behind Intel 45nm on pure transistor perf., if not quite ahead judging by their research experimental data we have from them.

KTE,

The ULK will have a positive power impact, but dynamic not leakage. Leakage in the backend is not the dominating power loss mechansims, gate leakage has surpassed subthreshold leakage at 65 nm.

On the Idsat, it will be interesting if AMD can implement the (110) technology since that will be a big boost in Pmos, they have not reported Pmos data that I am aware, though I have only 'lazily' looke dor it.

Their NMOS Idsat is not that impressive, they are reporting 1364 at 200 is about 8% above where they reported the initial 65 nm at the same subthreshold and this does not account for gate leakage which is only present during the on state. So, unless PMOS comes in very strong, then 8-10% more clock at the same thermals, or some power savings at the same clock.

It is hard to tell, or even speculate at this point since the initial 65 nm data is sorta stale, in that what AMD reported in 2005 was probably operative at 65 nm launch (which top binned at 2.6 GHz for a dual core at the time), but today that is pretty much different for what is going into Barcelona since they had a year to exercise CTI.

In short, I don't really see AMD's 45 nm really coming very close to Intels unless two things happen... they bring in HK/MG and come out with a very strong PMOS (which could happen if they do 110)

EDIT: I went and looked up the IEDM paper referenced in your link http://ieeexplore.ieee.org/xpl/freea...t=281&index=75 (oooops, not that one, this one http://ieeexplore.ieee.org/xpl/freea...t=281&index=78) and yep, their Ion/Ioff curve is very impressive.


Jack

[KTE]

KTE,

The ULK will have a positive power impact, but dynamic not leakage. Leakage in the backend is not the dominating power loss mechansims, gate leakage has surpassed subthreshold leakage at 65 nm.

On the Idsat, it will be interesting if AMD can implement the (110) technology since that will be a big boost in Pmos, they have not reported Pmos data that I am aware, though I have only 'lazily' looke dor it.

Their NMOS Idsat is not that impressive, they are reporting 1364 at 200 is about 8% above where they reported the initial 65 nm at the same subthreshold and this does not account for gate leakage which is only present during the on state. So, unless PMOS comes in very strong, then 8-10% more clock at the same thermals, or some power savings at the same clock.

In short, I don't really see AMD's 45 nm really coming very close to Intels unless two things happen... they bring in HK/MG and come out with a very strong PMOS (which could happen if they do 110)

Yeah Jack, the (110) PMOS is to be implmented at a later node because of its complexity but the ability has already been proven without even the need of HK/MG, some very good SOI performance has already been obtained as noted above. HK/MG will only add more benefits such as PMOS Idsat increase, after the 45nm fabrication process is matured, the integration will begin. I don't think we can judge PMOS performance based on NMOS performance data since Intels 45nm NMOS Idsat boost was only ~12% but PMOS was a huge ~50%. However the CMOS performance IIRC is not like Intels HKMG 45nm and without HKMG. The last research studies and papers on IBM/AMD SOI 45nm CMOS without HKMG is what I believe Shanghai will be based on, maybe a little tuned but not much more: PMOS 840μA/μm DC and NMOS 1240μA/μm DC at Vdd=1.0, Ioff=NA: http://hasan.nayfeh.googlepages.com/...ology_2006.pdf

[JumpingJack]

Yeah Jack, the (110) PMOS is to be implmented at a later node because of its complexity but the ability has already been proven without even the need of HK/MG, some very good SOI performance has already been obtained as noted above. HK/MG will only add more benefits such as PMOS Idsat increase, after the 45nm fabrication process is matured, the integration will begin. I don't think we can judge PMOS performance based on NMOS performance data since Intels 45nm NMOS Idsat boost was only ~12&#37; but PMOS was a huge ~50%. However the CMOS performance IIRC is not like Intels HKMG 45nm and without HKMG. The last research studies and papers on IBM/AMD SOI 45nm CMOS without HKMG is what I believe Shanghai will be based on, maybe a little tuned but not much more: PMOS 840μA/μm DC and NMOS 1240μA/μm DC at Vdd=1.0, Ioff=NA: http://hasan.nayfeh.googlepages.com/...ology_2006.pdf

Yeah, see my edit above. Si(110) is still in the research phase from my impression , simply built upon thumbing through articles -- I have not spent inordinate amounts of time reading every paper.

But looking at the experimental data ... it is quite impressive. The 1 uA/um Ion is taken at 100 nA/um Ioff is right there with Intel's current 45 nm.... the question is how close to moving from lab to fab is it? Not sure.

ITSA has not really published PMOS data, only one of two reasons a) they don't have anything to publish or b) it is so phenomenal they don't release the information for competitive reasons.

EDIT: Yeah the 45 nm data you linked here is kinda old data, in fact, it is showing about equivalent to current 65 nm process... this has been improved by now as seen by the NMOS data, so I take that 2006 info with a grain of salt (not the quality of the data, just the applicability).

EDIT2: Another thing caught my eye, you are right, PMOS and NMOS come down differently. The 50% gain you quote here is a massive jump related more to solving poly depletion problems by going metal. This is the largest node over node gain I have seen through IEDM. I think 50% is asking a lot from conventional poly (assuming no 110 -- that, again was impressive).

Edit 3: I went back to find the VLSI symposium data that yielded the NMOS numbers, it was leaked in Jan of 07, presented in june of 07. The Reg leaked it:

http://regmedia.co.uk/2007/01/28/ibmhighk.pdf

So even that NMOS data is old.

Jack

[madfaze]

1.4v KTE...

[bgbdbstrd]

Like many k9a2 + Phenom b2 owners, I have been requesting MSI to include the NB multiplier to a new BIOS release...here's the reply got today...

Dear Sir/Madam,

Thanks for contacting MSI Technical Support.

I am so sorry to tell you that BIOS engineer didn't succeed in adding this item.By testing,the PC always hangs if adding this item.So we can not release this BIOS.

So Sorry for any inconvenience to you.

Thanks and Best Regards
MSI Technical Support Team

Posted: Apr 01, 2008 by MSI


I asked them how come it was available in earlier bios versions and is/was instability the reason they took it out in succeeding ones....now i feel like i've got a hampered processor...and a motherboard...

[KTE]

1.4v KTE...

Which BIOS was this?

Like many k9a2 + Phenom b2 owners, I have been requesting MSI to include the NB multiplier to a new BIOS release...here's the reply got today...

I asked them how come it was available in earlier bios versions and is/was instability the reason they took it out in succeeding ones....now i feel like i've got a hampered processor...and a motherboard...

Yeah I figured this was the problem from earlier. I know this is why they took it down and this is why they made V2 but even that didn't correct it, it was a hardware fault. So if you tried 113 BIOS, although it did allow NB FID changing, it was very buggy and restricted clocking. I have not seen any board BIOS which works flawlessly with just these basics yet, they're having too many problems. here's hoping the nForce boards can get things much better, but true, I feel ya pain.

Yeah, see my edit above. Si(110) is still in the research phase from my impression , simply built upon thumbing through articles -- I have not spent inordinate amounts of time reading every paper.

But looking at the experimental data ... it is quite impressive. The 1 uA/um Ion is taken at 100 nA/um Ioff is right there with Intel's current 45 nm.... the question is how close to moving from lab to fab is it? Not sure.

From looking at Dr. Hasans research and papers (one of the chief scientists heading IBM/AMD 45nm technology) it looks to be a 32nm implementation and not until then, they don't have the mass tooling required either. They've gone past the first phase of research, process, initial product test but not far past that that I know of. It's certainly a future target.

ITSA has not really published PMOS data, only one of two reasons a) they don't have anything to publish or b) it is so phenomenal they don't release the information for competitive reasons.

c) could also be the exact opposite of b) as we saw with AMD K10h "hushness".

EDIT: Yeah the 45 nm data you linked here is kinda old data, in fact, it is showing about equivalent to current 65 nm process... this has been improved by now as seen by the NMOS data, so I take that 2006 info with a grain of salt (not the quality of the data, just the applicability).

I've just checked again, it seems I've got two or more papers mixed up from 2006 and 2007 on 45nm. The ones I initially read were of late 2007 detailing 45nm SOI CMOS results. Just checked and those values are the same as what ITSA claimed in 2006, but the 2007 DC NMOS Idsat is 1240 uA/nm.

EDIT2: Another thing caught my eye, you are right, PMOS and NMOS come down differently. The 50% gain you quote here is a massive jump related more to solving poly depletion problems by going metal. This is the largest node over node gain I have seen through IEDM. I think 50% is asking a lot from conventional poly (assuming no 110 -- that, again was impressive).

Agreed, no way SOI CMOS without HKMG would boost p-FET Idsat by 50% over, you'd need a "special" change for that such as HKMG. Usually we are talking +100-150 uA/nm at the same Ioff and Vdd maximum being a healthy increase down a full node.
I'll link you any updated 45nm SOI PMOS data once I become aware of them.

BTW check my 2x1GB 5-4-4-5 540 2.096v on P35 fully stable guys:



Awaiting to now compare how it goes on a Phenom/X2.

[KTE]

Like many k9a2 + Phenom b2 owners, I have been requesting MSI to include the NB multiplier to a new BIOS release...here's the reply got today...

Dear Sir/Madam,

Thanks for contacting MSI Technical Support.

I am so sorry to tell you that BIOS engineer didn't succeed in adding this item.By testing,the PC always hangs if adding this item.So we can not release this BIOS.

So Sorry for any inconvenience to you.

Thanks and Best Regards
MSI Technical Support Team

Posted: Apr 01, 2008 by MSI


I asked them how come it was available in earlier bios versions and is/was instability the reason they took it out in succeeding ones....now i feel like i've got a hampered processor...and a motherboard...

I've just talked with MSI Taiwan now about when I can have their latest NB FID/VID BIOS (etc) =>

Please wait for one that we're developing. Sorry for any inconvenience caused to you.

That's most likely *not* an official release.

[madfaze]

it is the latest official release bios from MSI version 1.4

[jonspd]

KTE yea if you could test some 2x2gb g-skill sets that would be great... also are you going to be testing some 9750 and 9850's?

[KTE]

madfaze: If I understand correctly, there's 4 major voltage domains on a Core 2 chip and for K10h, +4x PLL+1x IMC+1x HT but the HT voltage is the same as the FSB voltage in the C2 chips. The PCIe links are controlled by the I/O hubs in the RD790 chipset, so if you were to increase their voltage, it may very well increase PCIe volts too.

jonspd: hows your Phenom running, any updates?

G.Skill, I'll try but not until after I can get a Phenom. It'll be 9850, most likely, I'm not bothering with more I have duties to attend to now.

BTW, somethings going on - it looks like stores were price gouging initially but something else also.
9850 BE yesterday dropped price across all stores here even before release.

It was 176-201 before yesterday everywhere but by yesterday all of them dropped the price to 142-149, which is around $20-30 lower than Q6600 G0 here and $40-50 lower than the cheapest Q9300 is "on order" for.

That's also $40 lower than 9600/9600 BE were retailed here for until mid-Feb '08. The new price is pretty good for a 2,5G IMO since I do reckon there's higher chance of 2.8-3G stable on these than the B2.

[JumpingJack]

From looking at Dr. Hasans research and papers (one of the chief scientists heading IBM/AMD 45nm technology) it looks to be a 32nm implementation and not until then, they don't have the mass tooling required either. They've gone past the first phase of research, process, initial product test but not far past that that I know of. It's certainly a future target..

Yeah, when you posted that information it got me thinking... when D. Wang published his summary information from the 2007 IEDM (http://www.realworldtech.com/page.cf...1608222300&p=6) I missed the info about him taking the ITSA NMOS numbers from the VLSI Symposium in June. The 2007 IEDM proceedings have been published, but I could not find anything on current state of ITSA 45 nm, thinking I just simply passed over it I took the RTW data at face value.

So, prodding a little, I noted the footnote, looked up the VLSI paper and wolla .. D. Wang is quoting intial HK/MG data from the IBM paper I linked above -- to date we only know that AMD will not initially use HK/MG so that data is irrelevant in that regard. Weird also in that data was generated before last Jan. so using his information is terribly in error ... we don't know diddly squat, that data is simply too old, too much can change and is different.

Nonetheless, the (110) technology is extraordinarily interesting, especially in the SOI implementation -- as I understand from reading up on this, one must 'dig' a hole and selective grow (110) oriented Si to form the device. From the experimental data, they can push to near 1 mA/um which is just a hair shy of a MG NMOS device... this is impressive. If they do cut it in at 32 nm, together the HK/MG this will be something very special indeed.

Jack

[KTE]

BTW, there's more Athlon line CPUs set to release

65nm Brisbane (512KiBx2)
Athlon X2 5600+ G2, 76W 2900MHz
Athlon X2 4600+ (ee) G2, 45W 2600MHz

Athlon X2 4850e+ G2, 45W 2500MHz
Athlon X2 4450e+ G2, 45W 2300MHz
Athlon X2 4050e+ G2, 45W 2100MHz

[Sempron I think]
2100 G1, 65W 1800MHz
2100 G2, 65W 1800MHz
2200 G2, 65W 2000MHz

Has anyone tried visiting MSI Taiwan (Q&A) and been given a Virus attack caution by your AV? I can't get in to check their responses because of my AV stopping the link connection saying it's linked to downloading a Virus. I'm not going to risk it frankly.

This is the link to MSI Online Customer Service [CAUTION!! have your AV at full blast so any virus is caught by it]: http://ocss.msi.com.tw/index.php?mod=questions&dop=list

Let me know if any of you also get a virus alert on trying to access it or if its just my AV playing up (I doubt...)

Yeah, when you posted that information it got me thinking... when D. Wang published his summary information from the 2007 IEDM (http://www.realworldtech.com/page.cf...1608222300&p=6) I missed the info about him taking the ITSA NMOS numbers from the VLSI Symposium in June. The 2007 IEDM proceedings have been published, but I could not find anything on current state of ITSA 45 nm, thinking I just simply passed over it I took the RTW data at face value.

So, prodding a little, I noted the footnote, looked up the VLSI paper and wolla .. D. Wang is quoting intial HK/MG data from the IBM paper I linked above -- to date we only know that AMD will not initially use HK/MG so that data is irrelevant in that regard. Weird also in that data was generated before last Jan. so using his information is terribly in error ... we don't know diddly squat, that data is simply too old, too much can change and is different.

I had read the papers from the authors when they were presented so RW article was just a second source for me. I'll clarify this out more:-

Intel released their 45nm HK/MG (internal process codename is P1266 BTW) at IEDM 2007, in the paper entitled A 45nm Logic Technology with High-k+Metal Gate Transistors, Strained Silicon, 9 Cu Interconnect Layers, 193nm Dry Patterning, and 100% Pb-free Packaging - K. Mistry is the Manager of this technology node and the main presenter.

In it they mentioned:

Many challenges with high-k integration have included VT pinning, mobility degradation due to soft optical phonons, and poor reliability [1-3]. Metal gate electrodes not only eliminate the poly depletion effect, but enable high-k dielectrics by screening the SO phonons that cause mobility degradation

A key challenge was to simultaneously integrate high-k gate dielectrics, optimal workfunction metal gate electrodes and highly strained silicon channels. Transistors feature 160nm gate pitch, 35nm physical gate length, hafnium-based 1.0nm EOT high-k gate dielectric, dual workfunction metal gate electrodes, enhanced channel strain, ultra shallow junctions & nickel silicide.

PMOS performance is improved by using high-k+MG as well as by increasing the Ge content of the embedded SiGe to 30% from 23% (65nm [8]) and 17% (90nm [6]) and by reducing SiGe proximity to the channel. Drive currents are benchmarked at 1.0V, a low 100nA/μm IOFF and at 160nm contacted gate pitch. PMOS drive current (Fig. 10) of 1.07 mA/μm demonstrates 51% improvement over 65nm [8].
NMOS drive current (Fig. 11) is 1.36mA/μm, 12% better than our 65nm transistors [8]. The average drive current improvement over 65nm is 32% at the same voltage and IOFF despite scaled transistor pitch. These represent the best drive currents reported to date for 45nm technology at low IOFF.

Thats where we have the whooper figures from. If you compare Intels PMOS/NMOS IDSAT vs. IOFF curve, it's a huge increase from 65nm to 45nm. Gate leakage is reduced >25x for NMOS and 1000x for PMOS, and the IDSAT improvement figures reflect this positively.

IBM/AMD's 45m SOI CMOS figures were first released at IEDM 2006, in the paper entitled High Performance 45-nm SOI Technology with Enhanced Strain, Porous Low-k BEOL, and Immersion Lithography

In it they mentioned:-

The list of FET-specific performance elements includes enhanced dual-stress liner (DSL), advanced eSiGe, stress memorization (SMT), and advanced anneal (AA). The resulting PFET/NFET Idsat values, at Vdd of 1.0V and 45nm GR gate pitch, are 840μA/μm and 1240μA/μm respectively. The global wiring delay achieved with k=2.4 reflects a 20% reduction compared to k=3.0.

Later Fig. 14, shows the PMOS/NMOS IDSAT vs. IOFF curve showing IDSAT DC:
PFET = 800uA/um @ IOFF 200nA/um
NFET = 1140uA/um @ IOFF 200nA/um

Then at the VLSI Symposium 2007, Rhiga Royal Hotel – Kyoto, Japan – June 14-16; IBM/AMD's latest figures were revealed but for 45nm HK/MG, in the paper entitled High-Performance High-k/Metal Gates for 45nm CMOS and Beyond with Gate-First Processing. Lead members from the 3 MFGS I know that attended and presented on topics were Jim Farrell (AMD), Stephen Kosonocky (IBM) and Vivek De (Intel).

In there was mentioned:-

Fig. 6 shows a DC Idsat of 1240uA/um @ Ioff = 200nA/um with a corresponding AC Idsat 1364uA/um (non-self-heated) in the nFET HK/MG devices at 1.0V

By using an integration method that is consistent with traditional high thermal budget gate-first CMOS processing we have demonstrated improved short channel control with high-k/metal gate devices compared to SiON. The aggressive Tinv scaling enables Lgate scaling and results in the highest performing HK/MG nFET devices to-date for the 45nm technology and beyond. BE pFET HK/MG devices are shown with the lowest Tinv to-date providing a gate-first integration path for dual BE HK/MG CMOS technology implementation.

If you check the figures, much data of NMOS and PMOS is given but the ION vs IOFF curve of PMOS is skipped altogether.

So yeah, these last figures shown by RW in Jan '08 are not 45nm data we're seeing for AMD's Shanghai, but they will be improvments on the first figures I quoted from IBM/AMD's 2006 45nm CMOS as shown above instead. But I'd expect you to know that

The latest figures we know given by IBM/AMD were at IEDM 2007 in the paper (110) Channel, SiON Gate-Dielectric PMOS with Record High Ion=1 mA/um Through Channel Stress and Source Drain External Resistance (Rext) Engineering and I believe some newer figures are hinted in the paper SiGe Selective Epitaxy: Process Development for High Performance 45nm CMOS Technology but I can't get access to that paper yet, it was submitted at the ECS 2007 Conference though.

[jonspd]

Been great... No updates or anything just holding still for alittle while. Enjoying cod 4 for a while is that price USD? So you can get ahold of the 9850s already?

I was gonna update the chipset drivers but changed my mine as my b-day is less then 1 month away and looking forward to some kinda upgrade.