AMD Phenom II TWKR Black Edition spotted

[saaya]

i dont think amd or intel designed their process'es to run well on either low or high temps... and amd chips can run high temps, they just dont clock as well when running hot.

thats actually a hint towards the properties of the process and the cpu design... if a chip doesnt clock well when it runs hot on air, the chip is limited by heat and will most likely clock well under cold.

intels chips clock well on air, even at very high temps, so heat is not really limiting them or holding them back a lot, which also explains why they dont gain much when going subzero...

i think in the end this has a lot to do with processor design as well and not just manufacturing process... but anyhow, what does this have to do with this new phenom extreme edition chip? who cares how they build it as long as it clocks well

[saaya]

3 vs 4 issue wide is extremely little difference. And the key problem is x86/x64. Other CPU archs easily scale to 10 issue wide. So for AMD staying at 3 issue wide is not an issue by itself. Specially not without HT. I would say that the 4 issue part in Core 2 is less than 5% used. For i7 with HT on its higher but still...

i dont think its just 5%... they did quite a jump in performance when they widened it up... 5% average performance maybe, but in some cases its 15% at least i think...

and lets even say its only 5%... well how many extra transistors do you need? look at core duo and core 2 duo, the diference is minimal, and you get a 5% boost... for the same boost from cache you have to increase the die size by probably at least 25%... at least... and to get a 5% perf boost by clockpeed you have to increase the clocks from 3ghz to at least 3150mhz... a whopping 150-200mhz extra...

and its not like they have to redesign the entire logic, we are talking about quadcores here, so you rework one core and just copy it 4 times... and its not a complete rework, you just add an execution unit like the ones you already have, and then you just need to tweak the decoder and some other logic a little bit... i mean geez, wth... we are talking about the second largest chip maker here with dozens of chip engineers and all they did for the past years is copy and paste almost exactly the same design and just add some extra logic here and there... how amd could ever come up with a64 from athlonxp with this way of designing their chips is beyond amazing actually... :P

come on amd! get the whip out on your engineers and let them do some actual work and not just maintentance and process node shrinking!

[informal]

i worked at amd and heard the name before the launch and told all people at markom that had anything remotely to do with the processor naming thats its terrible and they should NOT use it... same for the logo... a comet? uhm a comet is a falling star, a star, something great and admired falling down, crashing and burning in flames and then exploding... hello? and thats what you markom people consider a good logo for a processor?

in the end it was the perfect logo for K10 cause K10 was amd falling down, crashing and burning and the cpus themselves were scorching hot

Just FYI ,a comet is not a star and therefore not a falling star .Look it up.You were probably thinking of a meteor(which is also not a star) or commonly known as a shooting star(logo is a comet and not a meteor).

[cdawall]

as chew* and sampsa have already said this chip does exist and it is only a quad core other than that it will retail a good bit above the 955 currently out the 42 does not stand for 4.2ghz it stands for 42 like the 42nd chip.

there will be some of these samples out shortly to a select group of people.

[Glow9]

i worked at amd and heard the name before the launch and told all people at markom

Shoot me a message if I can apply for an entry level marketing job at AMD cause I've been bashing my head with their naming scheme since the XP lineup. XP came out so snag that, then it seemed like Gillete was going hardcore marketing with Phenoms so AMD went with that. It's driving me nuts, but I agree I can't wait to hop back to AMD finally getting it together again. I'm GLAD the last CEO got the boot. All my hardware is AMD/ATI except this one rig and it's felt really weird having it and I am not even going to upgrade it I'm just going to hop back to AMD. But the name putting a "phenom" in my rig urghhhh curdling name..

[leonarderman]

I hope someone can post some result here a few day later!

"42" is a secret number, I think it's a 42th number

[eRazorzEDGE]

same for the logo... a comet? uhm a comet is a falling star, a star, something great and admired falling down, crashing and burning in flames and then exploding

just pointing out that a comet is a large body of ice and (possibly) rock that orbits a star once every so many years. take Halley's comet for example, comes around once every 75-76 years.

i believe you're thinking of meteor's, as in meteor showers. chunks of space debris that fall to earth and burn up in the sky.

*edit* lol, informal beat me to it while i was out looking at my broken into car... god damned theives

[informal]

just pointing out that a comet is a large body of ice and (possibly) rock that orbits a star once every so many years. take Halley's comet for example, comes around once every 75-76 years.

i believe you're thinking of meteor's, as in meteor showers. chunks of space debris that fall to earth and burn up in the sky.

*edit* lol, informal beat me to it while i was out looking at my broken into car... god damned theives

Haha sorry man,I was lucky you were busy

[Shintai]

i dont think its just 5%... they did quite a jump in performance when they widened it up... 5% average performance maybe, but in some cases its 15% at least i think...

and lets even say its only 5%... well how many extra transistors do you need? look at core duo and core 2 duo, the diference is minimal, and you get a 5% boost... for the same boost from cache you have to increase the die size by probably at least 25%... at least... and to get a 5% perf boost by clockpeed you have to increase the clocks from 3ghz to at least 3150mhz... a whopping 150-200mhz extra...

and its not like they have to redesign the entire logic, we are talking about quadcores here, so you rework one core and just copy it 4 times... and its not a complete rework, you just add an execution unit like the ones you already have, and then you just need to tweak the decoder and some other logic a little bit... i mean geez, wth... we are talking about the second largest chip maker here with dozens of chip engineers and all they did for the past years is copy and paste almost exactly the same design and just add some extra logic here and there... how amd could ever come up with a64 from athlonxp with this way of designing their chips is beyond amazing actually... :P

come on amd! get the whip out on your engineers and let them do some actual work and not just maintentance and process node shrinking!

How can you claim the performance is due to being 4 issue wide and not anything else? In the majority of code you barely use 2. Remember all the Hyperthreading block diagrams? Sometimes there is even only 1 (or 0) execution unit in action.



I would say most of core 2s performance over core comes via single cycle SSE and memory disambiguation plus improved prefecthers. And you can see how close Phenom 2 is to Core 2 in 3 vs 4 issue wide. However when AMD one day puts in hyperthreading I am sure they will change to 4 issue wide.

[azraeel101]

I still hate Phenom for the name.

I like it, when I hear Phenom I think of this:

[saaya]

Just FYI ,a comet is not a star and therefore not a falling star .Look it up.You were probably thinking of a meteor(which is also not a star) or commonly known as a shooting star(logo is a comet and not a meteor).

i know, but thats just arguing over small details...
lava is called magma before it comes to the surface, then its called lava...
meteorites or are called comets or asteroids before they enter the atmosphere...

the thing is, most people call it falling star, and thats a beautiful image, but NOT what you want people to relate to your product... a rock burning and crashing down in the sky :P

Shoot me a message if I can apply for an entry level marketing job at AMD cause I've been bashing my head with their naming scheme since the XP lineup. XP came out so snag that, then it seemed like Gillete was going hardcore marketing with Phenoms so AMD went with that. It's driving me nuts, but I agree I can't wait to hop back to AMD finally getting it together again. I'm GLAD the last CEO got the boot. All my hardware is AMD/ATI except this one rig and it's felt really weird having it and I am not even going to upgrade it I'm just going to hop back to AMD. But the name putting a "phenom" in my rig urghhhh curdling name..

no idea... amd is using interns to do cheap and basic work, like most big companies these days, you might get a position that way, though i doubt it... amd tend to do marketing externally afaik... smart thing really, cause in my experience markom people are the most likely employees to spend 90% of their day on facebook myspace and msn, if you pay them on a per project basis though, they NEED to keep moving

sure, there are exceptions to this... but ive seen this happen quite often...

shintai, well i dont know that much about cpu architecture at all... but what i do know is that amd barely improved their cores ipc, if at all, over several years, which is just not acceptable... yes they are "fast enough" for almost everything in daily computing, but the market obviously doesnt think the processors are fast enough, otherwise they would have a bigger share than 20% and you can find almost 0 highend pc with an amd cpu inside.

and its contradicting with their other strategies... if they really want to build a fast enough processor and dont even try to get their ipc up, why do they focus on quadcore cpus then? native quadcores even? cause even as of today most cpus sold are dualcores, and those really are FAST ENOUGH for most faily computing tasks...

and they obviously spent some time on deneb to tweak the inter core performance of their quadcores, which is nice, but i cant help but wonder, why you would spend such an effort on getting several underperforming parts to work well together instead of going for a fast redesigned part and just bolt several of those together without much hassle. which is btw what intel has been doing for quite a while, and quite succesfully i might add...

it still strikes me as ironic that intel came up with 2 cores glued together and not amd... cause its an ingenious economic solution in the original amd style... you can create 2 products ranges with one chip, and cover two segments with one original product... and if amd would have been smart they could have had quadcores (2 dualcore dies) at almost the same time they launched their dualcore athlonx2 in 2005, over 1 year before intel launched the core2 quad... sigh...

[Chad Boga]

the thing is, most people call it falling star, and thats a beautiful image, but NOT what you want people to relate to your product... a rock burning and crashing down in the sky :P

I don't see why you should rip AMD marketing for a rare moment of honesty.

[Tomasis]

bulldozer is cool name.. it is for to clean the rest of falling phenom haha

[Chad Boga]

bulldozer is cool name.. it is for to clean the rest of falling phenom haha

Well Sledgehammer/Hammer was also a cool name and they were great chips, let's hope that tradition continues.

[gosh]

i dont think amd or intel designed their process'es to run well on either low or high temps... and amd chips can run high temps, they just dont clock as well when running hot.

Is there any information (link) about how they clock on high temps?
It could be that they have set the temp limit low because they want a margin to avoid people break CPU's by overheating them.

[Smartidiot89]

Is there any information (link) about how they clock on high temps?
It could be that they have set the temp limit low because they want a margin to avoid people break CPU's by overheating them.

AMD has set a recommendation on max 1.5v for CPU/NB/HT voltages for overclocking, and a hardlimit at 62C - exceeding this often results in instability...

I have no link but these are words coming directly from AMD.

[[XC] gomeler]

They obviously refined the process for extreme cold OCing. Hense why it does so much better then Intel's at extreme cold. I highly doubt my chip is running higher then the temp reads on my 940. everything is cold (that is actually cold) to touch, board below chip, heatsink, etc. they are totally to different processes. get over it everyone.

AMD's lack of coldbug was a fluke. They didn't design for it, they simply realized during testing prior to launch.

[Lastviking]

Phenom II 940BE works fine @ 76c with 1.5v... i have tested higher temps without any harm and instabilty...anyone know what temp the cpu self trottel becuse of high temp?.. i have also done that one time...forget to plug in the fan , something was wrong when i runed 3dmark2006 low fps...hmm what is it and i open cpu-z hmm low mhz why?..hmmm something smells hot "ops the fan "

Intel works great at high temp...but who say that Amd dont do that also?.... but we dont need to have high temps becuse Phenom II dont geting hot like Core i7.

Main problem is weak pwc on the Amd boards.... they geting very hot,thats why its hard to test stabilty with high temp. My Dfi 790FXB did melt down something under the pwc area(under the board).

[xdan]

4ghz on stock voltage would be sweet (64bit)
i need 1.58v to get 4ghz semi-stable under 64bit

Quoted from a page of this post...
I think that ,this it's what Amd want's to improve
Not cold bug, not tolerance to high voltages.
I think that this "Tweker" will have a better scalability in overclocking with less volts , I mean it's a shame 1.58v for 4ghz
May be it's a new revizion improve like the notorius "E0" on Intel..
May be this will do 4.4-4.5ghz on 1.5v max... to equal Q9550/Q9650...
I bet that it will have 3.33ghz like I7 975, i mean it will be the top of Amd, so why shouldn't be equal at least in frecuency.... 3.4,3.5 seems like Prescott era .....

[AKM]

AMD has set a recommendation on max 1.5v for CPU/NB/HT voltages for overclocking, and a hardlimit at 62C - exceeding this often results in instability...

I have no link but these are words coming directly from AMD.

AFAIK the 62C figure is actually Tcase max not the actual core temps.

[Smartidiot89]

AFAIK the 62C figure is actually Tcase max not the actual core temps.

Sorry Tcase?

Afaik Phenom II has only 1 single tempsensor in the middle of the core, and not 1 separate per core, but maybe thats what you meent?

[gosh]

Intel works great at high temp...but who say that Amd dont do that also?.... but we dont need to have high temps becuse Phenom II dont geting hot like Core i7.

if you clock i7 is the L3 cache also clocked?
If I remember right then the L3 on i7 runs at the same clock as the cpu.
On Phenom the L3 cache runs on it's own clock

[cdawall]

Quoted from a page of this post...
I think that ,this it's what Amd want's to improve
Not cold bug, not tolerance to high voltages.
I think that this "Tweker" will have a better scalability in overclocking with less volts , I mean it's a shame 1.58v for 4ghz
May be it's a new revizion improve like the notorius "E0" on Intel..
May be this will do 4.4-4.5ghz on 1.5v max... to equal Q9550/Q9650...
I bet that it will have 3.33ghz like I7 975, i mean it will be the top of Amd, so why shouldn't be equal at least in frecuency.... 3.4,3.5 seems like Prescott era .....

just FYI voltages are different on AMD than intel what seems high on intel isn't high on AMD

[xdan]

just FYI voltages are different on AMD than intel what seems high on intel isn't high on AMD

I know the voltage story Intel vs Amd ..blah...
You didn't understood what i meant..
I meant to say that this "Tweaker" may have a better scalability frecuency vs voltage..
I mean a 940BE/955BE can do 3.6-37ghz stock voltages, and for 300-400Mhz for 4ghz you need to bump them 0.18-0.22v(from 1.35v stock for example)..
Wouldn't be great to have a cpu to do 3.6GHZ 1.35v, 4GHZ 1.4v, 4.2GHZ 1.45, and 4.5ghz 1.5v?
And as a note the previous 65nm generation from Intel had up to 1.5v tehnic specification like Amd .
So current 45nm AMD= 65nm on Intel alt least at voltages..
But even that it had some magic(one i saw in this forum) Q6600 with 1.15 vid. These Denebs are to damn linear(1.3v,1.35, 1.37) not one "magic" low vid. But at least i can say that AMD use a better manufacturing technology SOI, because all Phenom II 940/955 overclock almost the same, i mean at least Intel technology process skip some low vid chips in every category from Pentium Dual core to I7

[cdawall]

I know the voltage story Intel vs Amd ..blah...
You didn't understood what i meant..
I meant to say that this "Tweaker" may have a better scalability frecuency vs voltage..
I mean a 940BE/955BE can do 3.6-37ghz stock voltages, and for 300-400Mhz for 4ghz you need to bump them 0.18-0.22v(from 1.35v stock for example)..
Wouldn't be great to have a cpu to do 3.6GHZ 1.35v, 4GHZ 1.4v, 4.2GHZ 1.45, and 4.5ghz 1.5v?
And as a note the previous 65nm generation from Intel had up to 1.5v tehnic specification like Amd .
So current 45nm AMD= 65nm on Intel alt least at voltages..
But even that it had some magic(one i saw in this forum) Q6600 with 1.15 vid. These Denebs are to damn linear(1.3v,1.35, 1.37) not one "magic" low vid. But at least i can say that AMD use a better manufacturing technology SOI, because all Phenom II 940/955 overclock almost the same, i mean at least Intel technology process skip some low vid chips in every category from Pentium Dual core to I7

my 955BE albiet a special one from AMD did 4ghz @1.4v stable on DICE but 4.5ghz took 1.55v to get bench stable on DICE. it still depends on the chip and honestle volts don't matter to me my old 945BE ES chip did 4ghz air stable@1.6v and 4.9ghz benchable on DICE with 1.7v