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

[Sly Fox]

LoL, yeah i noticed that too, but i attributed it to people thinking jpg's are what they should use, instead of png's.



Well glad to know now, that those unbelievable results were just that, unbelievable ^^

Oh man, that comic is pure gold.

[informal]

Some interesting facts from here,may be of interest to some people:

AMD showcased several demo PCs using different cooling technologies with their new Phenom II X4 processor's on Gigabyte 790GX motherboards. The air cooling setup used a Cooler Master Hyper Z600 heatsink (running pretty quiet), the water/tec cooling had a CoolIT Freezone Elite, the phase change I *think* was a VapoChill, and best of all, an aluminum block for LN2.
Since these first Phenom II processors will be socket AM2+ only, at most a simple BIOS update should be all that is required to ensure full support / compatibility with existing boards.

Now, back to the presentation... During the LN2 demo, one of the first questions asked from the crowd was, "Is there a cold bug?"... Everyone got a good chuckle from that. Unlike the current 65nm Phenom's which don't seem to like extreme cold, there seems to be no issues with the Phenom II (booting or running) at very low temperatures. The probe temperature actually got lower than what is in the picture (that's the only picture AMD provided with a temp reading).

We are not allowed to give specifics on speeds at this time, but as a generalization one can expect the following stable performance without too much tweaking effort:

* Air: Around 4GHz is a safe area to target with decent aftermarket cooling & mild voltage increase. When I say "around" that means above and below, so don't gripe at me if your CPU won't hit 4GHz, there is no guarantee with overclocking. Also there aren't exactly huge amounts of CPUs out in the wild yet, so observations are somewhat limited.
* LN2: If you were thinking 5GHz, guess again (and guess higher)... It was like the Energizer Bunny commercials... they just keep going, and going, and going. To say in excess of 5GHz is an understatement, but that's about as detailed as I'm allowed. There was another large leap too between what was "stable" and what you could get a CPU-Z screenshot of, so that shows even more promise in the long term as yields will undoubtedly improve.
* Water & Phase Change: Obviously somewhere between Air & LN2. This is really going to vary based on what you are using and your temps

While I can't give exact numbers in terms of GHz, I can give you some of the voltages and temps seen while running the Crysis demo. Keep in mind these temps are not anywhere near 100% CPU usage, but they aren't exactly idle either.

* Air: Cooler Master Hyper Z600 - Around 1.5-1.55V / 33C
* Water/Tec: CoolIT Freezone Elite - Around 1.6V / 38C
* Phase Change: VapoChill - 1.7V / Below 0C (software / MB wouldn't report negative numbers)
* LN2: 1.75V (at least during the Crysis demo... *grin*)

It seems AMD will update AOD and Fusion utility to provide some basic auto OCing(my guess),probably in effort to match core i7s turbo -both, in my opinion, are just gimmicks ,but well...Here is the quote:,

The extra headroom that these new 45nm processors have give AMD the opportunity to really explore and expand their software overclocking tools. Around the launch of the Phenom II, look for updates to the AMD Overdrive utility and AMD's Fusion for Gaming application. These tools (along with the ATI Overdrive utility) really cater to PC Enthusiasts and Gamers of all skill sets. AMD has worked hard to make the interface as simple and automated as possible (for people that just want to game), while still featuring advanced menus for fine tuning (for people that really love to tweak every last bit).

All in all a great article!

Now on to some technical stuff,since someoen asked about IPC Deneb will bring.Well,it seems AMD gave the figures to german planet3dnowwebsite team:

* Approximately 3% come through improvements to the IPC standards.These come mainly through improvements in the Branch Prediction, the TLB and the pipeline.
* Approximately 10% of the larger L3 cache
* And again last about 5% through the transition from DDR2 to DDR3 memory.


Additional features will be smart Cool, application-specific profiles, an adjustable auto clock function and improved hardware monitoring with adjustable fan control.Details, but unfortunately there were none.

You can see what AMD lists in IPC improvement list.Seems DDR3 will bring approx. ~5% more perf. when compared to AM2+ Denebs.So overall,going to AM3 Deneb from AM2+Agena one gets ~18% advantage per clock.AM2+ Denebs should be ~13% faster than Agena if these numbers are true(but this is kinda overall gain;somewhere will be higher(pov ray ie. ~21%),somewhere lower (Nero recode~5%) ).Shanghai seems to confirm these numbers so it may very well be true for desktop too.

[Macadamia]

LoL, yeah i noticed that too, but i attributed it to people thinking jpg's are what they should use, instead of png's.

lol nice comic!

However if they use a proper imaging program like IrfanView or Photoshop, JPG saving at 80+% quality is extremely close to PNG, 90%+ almost artifact free.


God damn MS Paint

[tbone8ty]

nice find imformal! most ive heard so far

[cegras]

Those are pretty hard volts for 45nm though - 24/7 operation?

[informal]

Those are pretty hard volts for 45nm though - 24/7 operation?

Def. vcore for Shanghai is 1.35V ,so 1.6V is 18% above stock.And by the looks of it,chips run very cool even on air while OCed so it shouldn't be an issue.

[Frontl1ne]

Wow, thanks for all that info Informal. Deneb sounds to be shaping up very nicely. 33 degrees running Crysis on air. Damn.

[villa1n]

Wow, thanks for all that info Informal. Deneb sounds to be shaping up very nicely. 33 degrees running Crysis on air. Damn.

That is impressive. Although i would have to invest in another form of heater though

[spursindonesia]

Hmm, so it's quite an advantage using DDR3 in Deneb's platfrom, right ? Perhaps that's exactly the direction i should take regarding my future upgrade plan. Hopefully, by the Q3 2009, DDR3 price would be much more affordable, there would be some price cut and better stepping from both AMD and Intel, and the exchange rate of my national currency against the greenback would be better than the current state (it has dropped around 35% in just about 2 months )

[dinos22]

look at that socket clearance

i hate that type of socket though....so used to LGA
wish AMD would change FFS

[Macadamia]

That is impressive. Although i would have to invest in another form of heater though

An overclocked Pentium D 805 or Phenom 9850 on a cheap budget as a web machine or dedicated server?

BTW,
125W absolute worst case (probably not), 110W = my estimated normal case load of the 940.
On a very theoretical basis, assuming resistance is constant, power's proportional to voltage index squared.

1.6V
(1.18^2) * 125 = 174.05
(1.18^2) * 110 = 153.16400

1.55V
(1.15^2) * 125 = 165.3125
(1.15^2) * 110 = 145.47500

1.5V
(1.11^2) * 125 = 154.0125
(1.11^2) * 110 = 135.53100

Raising clocks will also raise power by a bit, so put that into consideration.
However as you've seen, 1.5-1.55 is VERY doable with a proper air heatsink. Even 1.6V can be done, but I suspect resistance issues start coming up from there- just my guess.

Frostytech has a NICE 150W test load, and you see the Z600 on it too.
http://www.frostytech.com/articlevie...id=2273&page=5

[BrowncoatGR]

Def. vcore for Shanghai is 1.35V ,so 1.6V is 18% above stock.And by the looks of it,chips run very cool even on air while OCed so it shouldn't be an issue.

Seems that way. The high default voltage must be because the very thick gates AMD seems to be using for 45nm. Iirc they only shrunk the gates 7% according to that EEtimes article.

[JumpingJack]

Seems that way. The high default voltage must be because the very thick gates AMD seems to be using for 45nm. Iirc they only shrunk the gates 7% according to that EEtimes article.

Actually, the 7% shrink is the gate length not gate thickness ... the gate thickness (oxide) was not mentioned.

[gallag]

If any cpu in the next 2 years runs at 4ghz @1.55v @ 33c on air cooling near full load I will eat my hat live on a web cam.

[informal]

If any cpu in the next 2 years runs at 4ghz @1.55v @ 33c on air cooling near full load I will eat my hat live on a web cam.

Read the article again.Crysis is,AFAIK,not stressing all 4 cores,hence it's not a full load.But yeah,you can rest assured these thing will run relatively cool(compared to other CPUs),even after heavy OCing and overvolting.Another big plus since nobody wants a melting socket/board or excess heat heating up other components.

[Rammsteiner]

If any cpu in the next 2 years runs at 4ghz @1.55v @ 33c on air cooling near full load I will eat my hat live on a web cam.

Read the article again.Crysis is,AFAIK,not stressing all 4 cores,hence it's not a full load.But yeah,you can rest assured these thing will run relatively cool(compared to other CPUs),even after heavy OCing and overvolting.Another big plus since nobody wants a melting socket/board or excess heat heating up other components.

lol reading skills

Oh, also, make sure you save that movie too! I want to show it later to my kids

[Macadamia]

If any cpu in the next 2 years runs at 4ghz @1.55v @ 33c on air cooling near full load I will eat my hat live on a web cam.

If you've seen my estimations coupled with the 150W synthetic load on LGA775 above you'd have some idea.

It depends on ambient temp. 15+18 = 33. More likely it would be 25+20 (more than 150W) = ~45.

[suddeath]

Another fake? However this one doesn't look like it's been shopped.

http://www.vr-zone.com/articles/phen...shot/6228.html

[Mads321]

Another fake? However this one doesn't look like it's been shopped.

http://www.vr-zone.com/articles/phen...shot/6228.html

Looks real enough to me (at least more real than the others).

Which got me thinking.. all the ppl doing fakes, you think they may be doing it to provoke ppl who has acutally got the real deal, to feel safer breaking NDA 'cause "loads of others already did it" (maybe even Intel employees?)

[Ghostbuster]

lol, The markbench guy has been exposed for his fraud on his own site..........look at comments on bottom of page http://markbench.blogspot.com/2008/1...very-step.html

There many more comments than before, and this one nails it on the head...

For english people: he said that has Windows XP SP3 during the tests, but on the scpeenshot on GPU-Z window you can see Vista 64 label.

[justapost]

Now on to some technical stuff,since someoen asked about IPC Deneb will bring.Well,it seems AMD gave the figures to german planet3dnowwebsite team:

Thank you for the info informal.

I read that few days ago in an german forum but here it is mentioned again.

For performance, we should because of a NDAS not much to say only this: against an Intel Core i7 965 beat us the available model excellent. We were on the ground to a Phenom II on a Gigabyte motherboard 790GX-hand and say goodbye to the statements of AMD employees to convince himself. A clock frequency of just under 4 GHz was on all test systems quickly and also achieved stable, all the benchmarks ran without problems through, whether 7-zip, wprime, or games like Crysis and Farcry2. Only the Cinebench R10 crashed on all available machines from reproducibly a given clock frequency, while all other tests still easily went through. What exactly is due, we can not yet say.

If it's not CB10 stable it's not prime95 stable.

[Macadamia]

"against an Intel Core i7 965 beat us the available model excellent"

What?


-----
Babelfish version!

To the achievement we may not say much due to a NDAs, only as much: against Intel a core i7 965 struck itself us the available the model outstanding. We were allowed to put on locally to a Phenom II on a gigabyte 790GX-Mainboard hand and to convince us by the statements of the AMD coworkers themselves. A clock frequency of scarce under 4 GHz was on all test systems fast and also stably reached, all bench mark went through without problems, all the same whether 7-zip, wprime, or plays such as Crysis and Farcry2. Only the Cinebench R10 fell on all machines the available reproducibly starting from a certain clock frequency, while all other tests went through still problem-free. On which to lead back exactly is cannot we at this time yet say.

So it's competitive AGAINST the i7 at 4 vs 3.2. They had to be convinced by AMD employees that it was?

[Ghostbuster]

"against an Intel Core i7 965 beat us the available model excellent"

What?

Translatation issues..

Will someone translate that into proper English?

[justapost]

I should have left out the first part. Here is a better translation.

Due to the NDA we can not say much about performance, but against an i965 our modell ran quite competitive. We where allowed to get our hands on an Phenom II running on an GBT 790GX board to prove this amd coworkers statement.

[AbelJemka]

If it's not CB10 stable it's not prime95 stable.

CB10 crashed on all machine it's not CB10 not stable.