AMD's Phenom II pushed to 6.5 ghz

[mr soft]

Wow this is an extreme OC. :scared:

http://es.youtube.com/watch?v=wB0JodKgZ0A

[Blue_MiSfit]

Wow :D

Very impressive indeed!!!

[Inventive Software]

Nice going! :D

[Dark Shikari]

They should keep going; if they get it to 7Ghz they might be able to compete with a stock Core i7!


;)

[Avenger007]

The insane clock speeds pale in comparison to the stability of the CPU near absolute zero using liquid helium! :eek:
Cold boot at -242°C, wow! :cool:

I wanna see the Core i7 do that with all it's complex power management circuitry and whatnot.
BTW, I haven't seen a [stock] Core i7 beat the world record in 3DMark 05 yet. ;)

[slow4mula]

that's cool and everything but anything you can't fun 24/7 is just for kicks...

[Sharktooth]

yep, it was just made to make a record.
a better version of that video is at this link (http://www.youtube.com/watch?v=wB0JodKgZ0A&fmt=18) btw.

[IgorC]

yes, Phenom II is extremely overclockable. There are some info http://news-amd.de/. AMD is planning to rise frequency up to 4 Ghz untill the end of 2010.
But there will be new technology 32 nm 4- and 8-cores and very cheap Core i7/i5 4 core+HT (8 virtual cores) from Intel by this time. Well, will see. More competition, more price cut. :)

I think most important advantage of Intel micros it's HT. While Both AMD and Intel can rise frequency as high as possible but AMD hasn't HT equivalent afaik.

[Sharktooth]

HT is not an advantage, it's a workaround to keep the pipeline filled.
It's like using 2 connections to download the same file when the speed is lower than the capacity of your line.
I dont think AMD processors would benefit from HT.

[IgorC]

I think it's advantage. HT gives 25% boost. http://www.fcenter.ru/online.shtml?articles/hardware/processors/25347

http://www.fcenter.ru/img/article/CPU/Phenom_II_X4/130975.png
AMD and Intel (wihout HT) are on par on x264 benchmark at the same frequency (more or less).

[Sharktooth]

it's not an advantage. its a "trick" for those CPUs that cant keep the pipeline full.
the core i7 is not faster coz of HT. as i said above, HT only helps keeping the pipeline full so all the units can work at full rate.
it's a way to ensure the CPU is always busy doing something instead of waiting for the pipeline to refill coz of too frequent cache misses or other issues.

[IgorC]

the core i7 is not faster coz of HT
You're hardly wrong here.
i7 w/HT 29.88 fps
i7 w/o HT 23.85 fps
HT brings +25% speed boost.

It's fact. Call it whatever you want advantage, disadvantage, whatever. But thanks to it Intel CPU are faster than AMD at the same frequency.
http://www.fcenter.ru/img/article/CPU/Core_i7/127162.png

HT only helps keeping the pipeline full so all the units can work at full rate.
Your statements have informative character and I've read articles about SMT/i7. There is no need to keep me in touch. It doesn't matter how SMT works but what impact brings for perfomance.

[Sharktooth]

you still dont understand.
HT doesnt speed up anything.
core i7 is fast but like P4 has issues with the pipeline...
it CANT keep it full... so HT is a workaround and is needed in some scenarios to keep the processor units running instead of waiting...
HT is not a magic wand that can make a CPU faster than it already is.
It just helps when it has troubles.
lets assume we have different algos that we run for test that should produce the same results in the same time but they're calculated in different ways.

case 1, ideal case. the i7 pipeline is always full and it executes an algo at "full speed". HT is not needed and in this case HT it may even hurt performance a little.

case 2, good case: the i7 pipeline gets sometimes stalled but thanx to HT it can execute an algo at almost full speed.

case 3, bad case: the i7 cant keep the pipeline full and in executes an algo inefficiently (slower) but enabling HT helps a lot and the execution speed is back to "almost full".

as you see in the example HT doesnt make the CPU faster, it is just an help in some circumstancies.
in other words, the CPU runs fast but you have to feed it with instructions to execute, otherwise it will wait for them... HT is there to ensure there are not "blanks" in the pipeline but it's a workaround to alleviate a problem in the prefetch or cache or someting.
theoretically the pipeline should be ALWAYS full, or there is something wrong in the CPU project and the instructions/data flow.
is that clear now?

[IgorC]

Now I understand you vision and way to express yourself. I have no problem with at all. :)
Let me rephrase myself in the case if I were you:
At the same frequency Intel CPUs are faster than AMD's at least for +25% in x264 benchmark.

End of discussion. I'm tired.

[Sharktooth]

i think the difference is even more with core i7 and intel has for sure room for improvement since core i7 needs a trick to work at almost full speed (and it doesnt always work - case 4 that is missing in my previous post). once they find what's wrong it will ALWAYS run at full speed and could be another step ahead of AMD...
however, i feel confident AMD will react. maybe not in the immediate future but soon enough to avoid bankruptcy.
it would be sad and counterproductive for the end user to not have competition.

[CruNcher]

Sharktooth
how should AMD go bankrupt positioning itself between Intel CPUs they can't :) (only if the gap between Price/Performance and Energy Efficiency becomes so huge it makes no sense anymore to buy AMD it will be problematic but that's yet not the case) and for a big target Group "Gamers" Intel CPUs make no sense @ all compared to the Price/Value, it's hard currently for AMD especially as they seem to put most R&D into fusion, but im also quiet sure i'ts much better for them to stay competitive against Larabee then wasting to much resources fighting Intel where they need no fight to be @ all :P
The only Company that could get problematic for AMD on high scale currently is Nvidia and a much Higher risk then Intel is for them currently and they very well technically position against them also the strategy to become a Service Partner of the Game industry combining Server knowledge and GFX (attack on TWIMTBP) is a sign of that.

[Manao]

Sharktooth : fullness of the pipeline isn't totally dependant on the CPU. The way a program is coded influences it greatly. Thus, HT is very good when you want to make your code as generic (cpu-wise) as possible.

[Sharktooth]

usually the compiler takes care of those cases (at least if it supports core i7 cpus).

[Astrophizz]

i7 just beat the Phenom II, and only at -50 C. http://www.bit-tech.net/blog/2009/02/03/intel-to-release-core-i7-975-new-stepping/

[Sharktooth]

i7 is not a budget CPU.
that said, we all know i7 is faster than phenom II, but the OP meant 6.5Ghz is an extreme overclock.
i7 at -50C was just at 5.2Ghz.

[Astrophizz]

Yeah, I dunno, I wouldn't call a $300 processor budget either, but that's in my eyes. I suppose I just meant to post a response to Avenger007's post (http://forum.doom9.org/showthread.php?p=1243491#post1243491), not much more.

[Shinigami-Sama]

Yeah, I dunno, I wouldn't call a $300 processor budget either, but that's in my eyes. I suppose I just meant to post a response to Avenger007's post (http://forum.doom9.org/showthread.php?p=1243491#post1243491), not much more.

more budget than 1000

[Avenger007]

i7 just beat the Phenom II, and only at -50 C. http://www.bit-tech.net/blog/2009/02/03/intel-to-release-core-i7-975-new-stepping/
It's interesting to see the Phenon II (http://service.futuremark.com/compare?3dm05=4696318) beat the Core i7 (http://service.futuremark.com/compare?3dm05=4729364) in the Firefly Forest test, although the difference is very small.
I just meant to post a response to Avenger007's post (http://forum.doom9.org/showthread.php?p=1243491#post1243491), not much more.
The point of my post was to emphasize the stability of the Phenom II chip at such an extremely low temperature; a feat which I have yet to see the Core i7 (or any Intel processor) repeat. The 3DMark05 comparison was BTW, but inevitable.

The aim of the experiment was to see how well the CPU scales at such extreme temperatures using liquid helium. Breaking the 6.5 GHz barrier from stock 1.8 GHz attests to the success of the experiment.

[Astrophizz]

Yes, good point.

[Sagittaire]

1) No phenon II is really not on par with core i7 ... lol
At the same frequency the phenon II is just on par with the old Q6600.

http://www.hardware.fr/medias/photos_news/00/24/IMG0024949.gif
http://www.hardware.fr/medias/photos_news/00/24/IMG0024954.gif

2) HT is really usefull for long pipeline like PIV. Anyway Phenon and Core use short pipeline (and Core i7 is by far more efficient than Phenon).
Well if HT is usefull for Core i7 certainely that HT should be really usefull for phenon.

[IgorC]

phenon
:rolleyes: Phenom

[Sagittaire]

:rolleyes: Phenom


Well phenon II or phenom II ... are just on par with the old QX6700 at 3 Ghz. Phenom II 920 at 2.8 Ghz is really good CPU for only ~180€. Anyway Core i7 920 is by far better CPU than Phenon II 940 for the equivalent price.

[IgorC]

It's really surprising that overclocked Phenom II 920 3.72 Ghz isn't faster than i7 920 without overclocking.
http://www.fcenter.ru/img/article/CPU/AMD_Phenom_II_X4_920/132165.png
http://www.fcenter.ru/online.shtml?articles/hardware/processors/26032

[Sagittaire]

It's really surprising that overclocked Phenom II 920 3.72 Ghz isn't faster than i7 920 without overclocking.
http://www.fcenter.ru/img/article/CPU/AMD_Phenom_II_X4_920/132165.png
http://www.fcenter.ru/online.shtml?articles/hardware/processors/26032

Not really big surprise. HT is really big advantage here for core i7 because x264 is highly optimized for threading (+25% with HT). Phenom II is just on par with old Q6xxx core 2. Anyway it's good performance because Q6xxx is really powerfull CPU. Core i7 is really incredible CPU and even phenom II with limit air cooling overclock (at ~3.9 Ghz) can't fight with him.

Personnaly I wait the new Q7500 with fsb at 800 Mhz. Certainely CPU with low price (~100€) and with the best overclock potential in the area (45 nm with fsb at 800).

[IgorC]

Wonder why Intel still doesn't sell Q7500. It could be amazing bomb if was here since 2008 (with E5/7xxx realese) to compete with cheap Phenom I.

[Sagittaire]

it's not an advantage. its a "trick" for those CPUs that cant keep the pipeline full.
the core i7 is not faster coz of HT. as i said above, HT only helps keeping the pipeline full so all the units can work at full rate.
it's a way to ensure the CPU is always busy doing something instead of waiting for the pipeline to refill coz of too frequent cache misses or other issues.

Well seem to be an advantage for the vast majority of application and particulary for massive threading application like video encoding. HT by only itself produce +25% for x264 with really efficient core i7 (core i7 is by far faster than phenom II at same clock even without HT).

HT is not here for resolve "something wrong in the CPU project" like you say but simply to optimize the pipeline work. IMO phenom II pipeline is certainely not always full and HT should be benefict too for phenom II.


Wonder why Intel still doesn't sell Q7500. It could be amazing bomb if was here since 2008 (with E5/7xxx realese) to compete with cheap Phenom I.

Certainely possible overclock at 4.2 Ghz in air cooling with cheap motherboard with 1333 Mhz bus support and better performance than phenom II (even with overclock at limit air cooling).

[CruNcher]

We shouldn't forget AMDs research got throwed back since the Athlon times (Financial Problems, restructuring with ATI) and the next big hit gonna be the Bulldozer Core from them though you can already estimate that by then Intels next Research cycle will be finalized too so practically AMD cant drive ahead anymore (buldoozer could become a comparable I7 now,unless they have developed something big in the labs in the meantime) but that's also not what they wan't currently they just try to position in a market between Intel to survive and concentrate their other resources into the fusion :)
We have a Company here whose constantly fighting to survive against 2 others, i have big respect in AMD especially as AMD is fighting without any tricks what you can't say about Intel's business strategies which weren't always whitehat ones in the past :P. Btw what they showed here looks for some just like a world record in a Game benchmark though for others it is a advancement in Material Research vs Intel :)

[Sharktooth]

Bulldozer is the next big step since Hammer.
Barcelona, Deneb, etc. were just small steps or "upgrades" to the Hammer core tech...
In an official announcement, AMD said Bulldozer will be its first substantially new CPU core since the original Athlon 64 processor of 2003. And it'll be the first ground up architectural redesign since the K7 Athlon of 1999.
So basically, Hammer tech is still competitive against Intel's Core2 tech... That's a big achievement from AMD.
They developed a core tech that last for 10 years...

[CruNcher]

Yep it will be a big thing for them and it could drive them ahead of Intel again maybe not for long but it would show the World that they still here and on the way to fusion, and im absolutely sure they gonna optimize it to work perfect in combo with a GPU every path, layer will be optimized for it :)

[Manao]

Sharktooth : from what I read, P3 -> PM -> Conroe / Penryn were quite close to one another too. And I don't see why making something last long is a good achievement. It can also show they didn't have the means to renew their architecture.

[Sharktooth]

just because they made a 30+ stages pipeline in the P4 it doesnt mean the p3 and conroe are the same tech. they just share some principles.
what i meant is AMD built an architecture that lasted for several years and with minor upgrades has still competitive performance.
bulldozer, btw, will feature thread-fusion, massive multithreading a new level of virtualization and things we "mortals" are not meant to know right now (or maybe those are just rumors). it's a complete redesing so we can expect it to be a completely new core tech that will have almost nothing in common with the previous one.

[Sagittaire]

just because they made a 30+ stages pipeline in the P4 it doesnt mean the p3 and conroe are the same tech. they just share some principles.
what i meant is AMD built an architecture that lasted for several years and with minor upgrades has still competitive performance.
bulldozer, btw, will feature thread-fusion, massive multithreading a new level of virtualization and things we "mortals" are not meant to know right now (or maybe those are just rumors). it's a complete redesing so we can expect it to be a completely new core tech that will have almost nothing in common with the previous one.

Well AMD promise always spectacular upgrade but in the fact phenom is not really impressive. IMO Intel have really confortable technology advance and for long time now. The most important part of inovation for multicore is low TPD control. The most spectacular inovation for me is ... Atom. With this technology Intel is ready for massive multicore CPU: for example CPU with i7 master core and 16 secondary atom core will be a killer for Video Encoding with TDP < 100 Watts

[Sharktooth]

i7 is just a plain ripoff of the AMD design. they copied and improved what AMD did several years ago. sure, it's damn fast, but nothing new from the design perspective.
atom, however, is the real "innovation" brought by intel. i believe they will use the atom core tech in several products (starting from larrabee).
however, we're going completely OT.

[Dark Shikari]

i7 is just a plain ripoff of the AMD design. they copied and improved what AMD did several years ago.What are you talking about? If you're referring to the nonsense about being a "true quad core," the benefit from that is quite tiny, perhaps a bit of L2 cache coherency. The real benefits of the i7--just like the real benefits of the Phenom--were not "stolen", they were the work of the people they pay a whole lot of money to make good chip designs. Any claim to the contrary is pretty much just stupid fanboy ranting.

Using the same logic, you can claim all sorts of absurd things, like that the Phenom is a ripoff of the Core 2, since they stole the idea of a fast 128-bit SSE execution unit.

[Sagittaire]

What are you talking about? If you're referring to the nonsense about being a "true quad core," the benefit from that is quite tiny, perhaps a bit of L2 cache coherency.


It's true that the old Q6xxx design (2x2 cores) is more efficient than Phenom II design.

[Sharktooth]

Dark Shikari, im talking about the integrated memory controller, the point to point bus link (hypertransport on AMD -> quickpath on Intel), 3 levels of cache (last one shared)... they even renamed x86_64 to EMT64... this reminds me the h.264 logo in quicktime as if they were the author of h.264...
should i continue? all those techs were already present on AMD chips.
The 128 bit SSE execution units is just a good idea, but not a revolution (from the chip design perspective).
As i said, Intel copied and improved... and they're quite good at improving.
So, the question is, who's the author of the REAL innovation?

[Dark Shikari]

Dark Shikari, im talking about the integrated memory controller, the point to point bus link (hypertransport on AMD -> quickpath on Intel), 3 levels of cache (last one shared)... they even renamed x86_64 to EMT64... this reminds me the h.264 logo in quicktime as if they were the author of h.264...
should i continue? all those techs were already present on AMD chips.
The 128 bit SSE execution units is just a good idea, but not a revolution (from the chip design perspective).
As i said, Intel copied and improved... and they're quite good at improving.
So, the question is, who's the author of the REAL innovation?There's no point in arguing with someone who knows so little about the process of chip design and the history of AMD and Intel and their cross-licensing agreements.

I'm not going to bother. Do the research yourself if you want to, but at least do everyone a favor and stop ranting about things you don't know about.

[Sagittaire]

Dark Shikari, im talking about the integrated memory controller, the point to point bus link (hypertransport on AMD -> quickpath on Intel), 3 levels of cache (last one shared)... they even renamed x86_64 to EMT64... this reminds me the h.264 logo in quicktime as if they were the author of h.264...


Come on ... hypertransport, 3 level of cache and 64 bits intruction are not real innovation.


they even renamed x86_64 to EMT64

Certainely that if Intel will be the first to introduce EMT128 then AMD will use x86_128. Where is the copy ... 64 bits instruction is natural evolution like all your (bad) other example. Why not speak to MMX, SSE, SSE2, SSE3, SSE4, SSE5 ... in this case.

[Sharktooth]

Come on ... hypertransport, 3 level of cache and 64 bits intruction are not real innovation.
sure... not innovations... sure...
oh you forgot the integrated memory controller... that's not an innovation too... sure...
Certainely that if Intel will be the first to introduce EMT128 then AMD will use x86_128. Where is the copy ... 64 bits instruction is natural evolution like all your (bad) other example. Why not speak to MMX, SSE, SSE2, SSE3, SSE4, SSE5 ... in this case.
MMX... pentium 1... why dont we talk about the 8086? :p
AMD also introduced instruction sets... they were 3dnow and 3dnow enhanced. and they were not ripoffs of SSE...
SSE5 is also from AMD...
however, as i said, we're going wildly OT... the intent of this thread was about extreme overclocking ability at extremely low temps of the phenom 2. so, intel fanboys, please calm down and come back when you have something to run at at least -240C...

[ash925]

Well in case of IMC Intel was first with Timnah.

[Sharktooth]

yeah the CPU that never existed...
please...

[lucassp]

So, the question is, who's the author of the REAL innovation?

I think DEC Alpha had an IMC and something similar to HT/QP.

[Shinigami-Sama]

I think DEC Alpha had an IMC and something similar to HT/QP.

to bad they killed the alpha :<

[STaRGaZeR]

I thought Doom9 was free of fanboys...

[froggy1]

I thought Doom9 was free of fanboys...

not when Sagittaire is around :p

[STaRGaZeR]

not when Sagittaire is around :p

I wasn't thinking about him precisely, but he's a fanboy indeed. Sharktooth wins. When I read his posts in this thread I think: does this guy really believe what he's writing? ...

[Sharktooth]

I think DEC Alpha had an IMC and something similar to HT/QP.
Most DEC engineers behind Alpha processors were hired by AMD in the old K7 days (another example was the EV6 bus).
So, basically, the IMC idea came from the same minds.

@stargazer: yes, i believe what i write and instead of blowing air out of your mouth, why dont you tell us something more interesting than insults?

@all: Just a reminder... again: the thread is about overclocking at insane low temps. some ppl here are hijacking this into "i7 is faster than Phenom2"... WE ALL KNOW THAT.
can we please get back on track? thanks.

[STaRGaZeR]

@stargazer: yes, i believe what i write and instead of blowing air out of your mouth, why dont you tell us something more interesting than insults?

Because of what DS said. You're so brainwashed that the fastest method would be a new brain, but since that's impossible it's best to not say anything and leave you in the ignorance you're now. I'm really tired of hearing the same nonsense coming from every good old AMD fanboy out there.

[Sharktooth]

and im tired of an intel fanboy that uses his mouth only for insults and hijacking threads.
JFYI (and for Dark Shikari's too) read: http://www.amd.com/us-en/assets/content_type/DownloadableAssets/AMD_-_Intel_Litigation_History.pdf
and now please get back on topic.

[neuron2]

Closing thread as the insults have eclipsed any useful on-topic content. I should issue a bushel of strikes but I'll settle for reminding you all about rule 4.