AMD TWKR Black Edition Hands-On

I'd like to know how they "hand pick" these processors in the first place. Anyone know? Do they overclock them themselves at the factory hoping not to burn out a few 1000 of them just to get a couple dozen that look promising?

If they've already overclocked the shit out of them to get these golden samples then what's the point of some other dufus doing the same thing later and reporting his findings?

I guess I don't get how the process works to finding these miraculous chips that overclock so much better than the chip that was cut from the same die right next to it.
 
Valthar said:
I'd like to know how they "hand pick" these processors in the first place. Anyone know? Do they overclock them themselves at the factory hoping not to burn out a few 1000 of them just to get a couple dozen that look promising?

If they've already overclocked the shit out of them to get these golden samples then what's the point of some other dufus doing the same thing later and reporting his findings?

I guess I don't get how the process works to finding these miraculous chips that overclock so much better than the chip that was cut from the same die right next to it.
Click to expand...

I am sure that AMD has better/more advanced testing equipment than you could think of that will tell them about these said miracle chips performance...
 
I wouldn't be surprised if they had a test bed that automatically scaled voltage and clock speed till if failed. Probably test a few out a batch and earmark them.
 
LibertySyclone said:
I am sure that AMD has better/more advanced testing equipment than you could think of that will tell them about these said miracle chips performance...
Click to expand...

That's what I'm asking... anyone know how they do this? I don't follow much about the details of fabricating chips and just wondered is all. AMD/Intel... hell even GPU's... just wondering how they tell that any particular chip is destined to be an "extreme" or "xxx OC" chip.

Thanks
 
Valthar said:
That's what I'm asking... anyone know how they do this? I don't follow much about the details of fabricating chips and just wondered is all. AMD/Intel... hell even GPU's... just wondering how they tell that any particular chip is destined to be an "extreme" or "xxx OC" chip.

Thanks
Click to expand...

No idea but it would be interesting to find out. My guesses involve setting a clock speed then dropping voltage until a chip is unstable then upping it a bit and selling it at the slightly upped voltage to cover margin of error.
 
MetalDwarf said:
I would love to see some home brew overclocking. i miss the days of hardware store watercoolers, and beer fridge phase change. it would be fun to see if you can find an inexpensive, easily available and modable phase change setup like a beer fridge for sale at walmart so anyone could buy it, rip it apart, and put it back together into a super cooler.

Ya same here. I still have my hardware store water cooler from the K6 days. Cut off the end of a plumbing cap and silicon it to the cpu plate. Water ran directly over the plate itself. I actually found the guys website on how to do it from a link here on the H A long time ago.Around 98 i believe. He would take his and put the water tank outside in the winter lol. I ran my k6 350@500 for a year with that.
Click to expand...
 
Valthar said:
That's what I'm asking... anyone know how they do this? I don't follow much about the details of fabricating chips and just wondered is all. AMD/Intel... hell even GPU's... just wondering how they tell that any particular chip is destined to be an "extreme" or "xxx OC" chip.

Thanks
Click to expand...

Google for AMD's "APM" technology. Or rather, GlobalFoundaries' APM technology now. I'm sure Intel and TSMC have the same types of automated feedback controls it's just called something different.

The whole idea behind running special "bins" inside of fabs is to get some chips made with certain characteristics that lend themselves to low power or high clocks or better yields, it's getting the right mix of all those consistently that makes or breaks the fab. These chip bins all use the same core manufacturing design but there are certain tweakable elements that you can control. This enables you to run a single fab line with a special mix of elements to yield an ultra low voltage CPU with tight leakage, low heat, and low clocks, while the next line over is producing your black edition parts that will never need to fit inside of those constraints.

For the record, "high leakage" isn't necessarily a bad thing, it just depends on what you're going for. The best example I can think of are the 2900XT parts. Those things leaked like crazy and ran hotter than hell because of it. You had to run it with high voltages (= high heat output) in order to get respectable speeds that were stable. If you tried to oc the stock part you probably didn't get that far and ran into thermal boundaries thanks to the leakage. BUT, if you could get the temps down enough so leakage didn't cause issues, that thing had some serious legroom in the clockspeed department and you could continue to add voltage (sounds a lot like the average Phenom II oc experience hmm). IIRC, extreme LN2 and dry iced 2900XTs held various clockspeed and 3DMark records for quite awhile, even after the 3870 was released on a much improved fab process, though with a neutered memory bus. The 3870 was designed for yields and energy efficiency (basically for mass market profit). The core design for the 2900 wasn't much different besides the obvious die shrink, but when it came to producing them the R600 had very different power and clockspeed characteristics than the RV670.

AMD's been at this for awhile so they know what the right mix is to produce a wafer that *should* yield chips that oc better than others. APM helps them out with that by giving them measurable details about what exactly causes the chips to be better for oc'ing. So while much of the Phenom II line is probably geared towards higher clockspeed scaling in general since AMD marketing really tries to hit on that, these TWKR chips are the best of the best. I'd assume they do have to hand pick and validate these chips, but it wasn't like the guys were looking for needles in a haystack. They knew the exact lot numbers, wafer id's, etc where these chips would be coming from and so only had to test a few of them before finding some cores with really fantastic results.
 
Silent.Sin said:
Google for AMD's "APM" technology. Or rather, GlobalFoundaries' APM technology now. I'm sure Intel and TSMC have the same types of automated feedback controls it's just called something different.

The whole idea behind running special "bins" inside of fabs is to get some chips made with certain characteristics that lend themselves to low power or high clocks or better yields, it's getting the right mix of all those consistently that makes or breaks the fab. These chip bins all use the same core manufacturing design but there are certain tweakable elements that you can control. This enables you to run a single fab line with a special mix of elements to yield an ultra low voltage CPU with tight leakage, low heat, and low clocks, while the next line over is producing your black edition parts that will never need to fit inside of those constraints.

For the record, "high leakage" isn't necessarily a bad thing, it just depends on what you're going for. The best example I can think of are the 2900XT parts. Those things leaked like crazy and ran hotter than hell because of it. You had to run it with high voltages (= high heat output) in order to get respectable speeds that were stable. If you tried to oc the stock part you probably didn't get that far and ran into thermal boundaries thanks to the leakage. BUT, if you could get the temps down enough so leakage didn't cause issues, that thing had some serious legroom in the clockspeed department and you could continue to add voltage (sounds a lot like the average Phenom II oc experience hmm). IIRC, extreme LN2 and dry iced 2900XTs held various clockspeed and 3DMark records for quite awhile, even after the 3870 was released on a much improved fab process, though with a neutered memory bus. The 3870 was designed for yields and energy efficiency (basically for mass market profit). The core design for the 2900 wasn't much different besides the obvious die shrink, but when it came to producing them the R600 had very different power and clockspeed characteristics than the RV670.

AMD's been at this for awhile so they know what the right mix is to produce a wafer that *should* yield chips that oc better than others. APM helps them out with that by giving them measurable details about what exactly causes the chips to be better for oc'ing. So while much of the Phenom II line is probably geared towards higher clockspeed scaling in general since AMD marketing really tries to hit on that, these TWKR chips are the best of the best. I'd assume they do have to hand pick and validate these chips, but it wasn't like the guys were looking for needles in a haystack. They knew the exact lot numbers, wafer id's, etc where these chips would be coming from and so only had to test a few of them before finding some cores with really fantastic results.
Click to expand...

That doesn't make sense at all. HD 4890 is less leaky than HD 4870 and with the same voltage it can reach a higher clock speed. PII is less leaky than original Phenom so PII can reach a higher clock speed using the same voltage. You can only increase the voltage until a certain limit before the chip will be damaged. If you look around XtremeSystems, you will see that the latest PII overclock will need less voltage to reach the same 6GHz clock speed under LN2 when compared to earlier PII. The latest overclocking event showed that they only need 1.475V to reach 6.8GHz
http://hothardware.com/Articles/AMD-Phenom-II-Hardcore-Overclocking-Event/?page=2
 
alg7_munif said:
That doesn't make sense at all. HD 4890 is less leaky than HD 4870 and with the same voltage it can reach a higher clock speed. PII is less leaky than original Phenom so PII can reach a higher clock speed using the same voltage. You can only increase the voltage until a certain limit before the chip will be damaged. If you look around XtremeSystems, you will see that the latest PII overclock will need less voltage to reach the same 6GHz clock speed under LN2 when compared to earlier PII. The latest overclocking event showed that they only need 1.475V to reach 6.8GHz
http://hothardware.com/Articles/AMD-Phenom-II-Hardcore-Overclocking-Event/?page=2
Click to expand...

The RV790 was a chip designed specifically around getting higher clockspeeds, they actually modified the power distribution circuitry to allow for it. It was physically designed to reach certain goals that couldn't otherwise be reached by speed binning. I wasn't trying to say that in order to get the most headroom out of your chip you should design it to be leaky to begin with. I don't think that was on any of the designers' minds when they were making the R600 or Barcelona. I was just saying that isn't necessarily a bad thing IF you are an overclocker going for higher potential clockspeeds under ideal conditions (extreme low temps). Yes, you can only increase the voltage to certain limits without damaging the chip, but that limit is usually closely related to the temperature of the chip so with better cooling the higher the safe voltage.

What I was referring to in relation to this TWKR chip is where the same exact die will react differently depending on different manufacturing qualities. There are still limitations of process nodes and design that you can't really get around (Phenom I) which determine the clockspeed ceiling. A poor IMC design or slow HT link can stop you in your tracks just as easily as a safe voltage limit. I'm sure you've read about or run into it yourself where even when adding voltage to a CPU it simply won't clock higher. As the process matures you should definitely expect to reach higher clockspeeds at the same or lower voltages, if you don't then something is seriously wrong with your fab or design. The fact that max PII oc's are happening at lower and lower voltages should be expected.

Take a look at this graph from GlobalFoundries that is a result of their APM technology. It can explain the relationship between wanting higher performance at the sacrifice of lower yields (and other factors) and finding the right balance between performance and yield. Going for balls out performance is going to reduce the number of chips that can run within certain voltage specs or work at all. That is the region where these TWKR parts are coming from. The R600 was aimed at that space because it was a top tier part and it underperformed compared to the competition so they had to try and get the most clockspeed they could out of it, similar to Barcelona. Higher performance isn't the result of leaky chips, leaky chips are the result of aiming for higher performance.
 
Has anyone actually figured out what the TWKR is?, in terms of clock speed or TDP or any of that?
 
This is the best article from FUD ever:
http://www.fudzilla.com/content/view/14455/1/
The '42' in the designation means absolutely nothing. It's a reference to Douglas Adams' Hitchiker's Guide to the Galaxy, in which a fictional supercomputer named Deep Thought works out 42 is the Answer to the Ultimate Question of Life, the Universe, and Everything after 7.5 million years of calculations. (Somewhat like trying to do GI in Max a few years back.sub.ed.)

AMD claims the cherry picked TWKRs can handle more voltage than death row inmates, scale higher and run faster than any processor on the market, and we could even see one of them break the 7GHz mark with liquid helium.
Click to expand...
 
ocellaris said:
Just further attempts by AMD to confuse the shit out of people and squeeze out additional chip SKUs from their limited manufacturing range. Nothing to see here folks.
Click to expand...

Are you saying these are going to be for sale?
 
O-P-M said:
Are you saying these are going to be for sale?
Click to expand...

Actually that reply appears to be an intel fanboy doing his thing...... And it looks like the answer to your question would be no if you check the link above you. :(
 
Yeah I knew that they weren't.

Appears Tom's Hardware is giving away 2 of them though.

You've seen the leaked photos of AMD's latest overclocking darling, the Phenom II X4 42 Black Edition TWKR. We got our hands on a pair of these CPUs (four, actually, but two are to give to you) and benchmarked them on a liquid nitrogen-fueled test bench.
Click to expand...
 
damn this could get interesting, with a sub 450$ price tag this could end up in my computer
 
Daeldus said:
damn this could get interesting, with a sub 450$ price tag this could end up in my computer
Click to expand...

LOL, if AMD lets them go public they already said $999 MSRP. This means fail to me even if they hit 4.5ghz on average. Course I hope the $999 MSRP is a distasteful joke on someones part but thats the only price I've seen attached to these yet.
 
Daeldus said:
damn this could get interesting, with a sub 450$ price tag this could end up in my computer
Click to expand...

LOL, try sub $400 if they manage hit 4.5ghz+ on average with 4.5ghz being the min expected. Someplace I saw that was going to be giving 2 away said $999 MSRP. This means fail to me even if they hit 4.5ghz on average. Course I hope the $999 MSRP is a distasteful joke on someones part but thats the only price I've seen attached to these yet.

With 4ghz almost expected with decent cooling on just about any PII this CPU needs to first hit 4.2 - 4.5ghz on good air and up to 4.7 - 5ghz on water to warrant anything higher than $400 imho.

Matter of a fact, these need to come out of the box @ 4ghz.

K, after actually reading the above reviews this CPU should be labeled EPIC FAIL. "This cpu is meant to be used under extreme cooling" kinda did it in. WELL NO SHIT YOU HIT 4.6ghz under extreme cooling dumbass. But it takes less volts to do it. Omfg, I bet I could grab a reg PII on an good mobo with good ram and hit 4.6ghz under dry ice. "But omfg, I broke 5ghz under LN2" . . . . . you and about 100 other people running reg PII's under LN2.

AMD, you fail me again =.=

If AMD wants to impress us they had better box up a PII that comes out of the box @ 4ghz with a proper HSF fast. They could simply make a Phenom II Heatpipe Tower Edition and salvage the PII as a contender if they can guarantee a 4ghz out of box PII with a heatpipe tower. This would be worth a $100 premium over a PII 955.

The PII has serious potential with a high enough clock out of the box but all this pidly conservative shit is killing them.
 
Last edited:
This is not a water chip or a chiller chip to get the max from it you need extreme vapor cooling.
-150 or greater to get those big OCs
 
d50man said:
This is not a water chip or a chiller chip to get the max from it you need extreme vapor cooling.
-150 or greater to get those big OCs
Click to expand...

All the more reason it shouldn't even exist. OverClockers in general make up less than 10% of PC users worldwide. Extreme OverClockers make up maybe .5% of that already less than 10%. Why even waste the time and resources marketing a product to .05% of all PC users worldwide?

I'm sure AMD did it to be able to go "Look what we can do" meanwhile Intel is laughing going "AND ITS STILL SLOWER THAN OURS ROTFL".

Worry about shrinking the clock for clock defecit you have against Intel's C2Q series first ( then the defecit you have on the i7's ) then you can worry about uber clocking CPU's AMD.

Now, if you'll excuse me I'ma go make a Michael Jackson love doll aimed towards kids. Because you know how theres such a big percentage interested in that and how its not waste of time and resources at all.
 
Last edited:
shadow_419 said:
Did anybody actually follow the link? These are not for sale and have never been intended for retail channels.
Click to expand...

In that case this is the dumbest ass idea AMD has had to date. "Lets release a CPU with massive potential then tease the public with it with no intent to ever make it a production model ROTFL."

Sad that they have this much time to spend cherry picking CPU's that will never reach the consumer yet they have no time to make a CPU actually competetive to Intels now generation old CPU's. I could maybe understand this if they were just simply behind the i7 but FFS they are now behind the Core 2 series with thier latest chips.

Everyone applaud AMD for this latest waste of time.
 
Why is it dumb?

1. Bin 100 or so cpu's
2. Give them to extreme cooling overclockers (just like ES)
3. They overclock the cpu's to speeds no intel quad can reach
4. ???
5. Profit from exposure
 
shadow_419 said:
Why is it dumb?

1. Bin 100 or so cpu's
2. Give them to extreme cooling overclockers (just like ES)
3. They overclock the cpu's to speeds no intel quad can reach
4. ???
5. Profit from exposure
Click to expand...

K, point 2 is null and void as they could have released these as ES but instead give them a special name only to make it mean nothing to the average consumer as they will never see them personally.

Point #3 is null and void because obviously you didn't look into extreme overclocking before you made that comment because there were people to hit 5ghz+ on C2D's looong ago so you're just plain wrong.

Point #5 makes no sense because . . . wtf? Negative exposure? Just because you give a CPU a special name and only release it to people that use extreme cooling means that you are a fn moron to have let that idea make it past the "List of ideas I had while waking up" stage.

Intel could easily do the same thing with thier regular CPU's and they'd get the same minimal exposure that will not make an impact on thier proffit . . . EVER. Why? 99% of CPU sales are in the server/work station market ( which doesn't OverClock ) or the average consumer market which doesn't know about or doesn't care about OverClocking or are just too scaered to.

The only people that will really see the TWKR CPU's and understand thier purpose are here and places like this and tbh it kinda dissapoints me to hear about a CPU that unlike its hand picked memory and VGA cousins will never be within my grasp.

Memory companies do this but they release thier hand picked parts to the public like ATI did with those pimpin highly OC'd cards in the sexy brief cases a few yrs back.

AMD = pissin me off with this publicity stunt.
 
Hard and Confused said:
K, point 2 is null and void as they could have released these as ES but instead give them a special name only to make it mean nothing to the average consumer as they will never see them personally.

Point #3 is null and void because obviously you didn't look into extreme overclocking before you made that comment because there were people to hit 5ghz+ on C2D's looong ago so you're just plain wrong.

Point #5 makes no sense because . . . wtf? Negative exposure? Just because you give a CPU a special name and only release it to people that use extreme cooling means that you are a fn moron to have let that idea make it past the "List of ideas I had while waking up" stage.

Intel could easily do the same thing with thier regular CPU's and they'd get the same minimal exposure that will not make an impact on thier proffit . . . EVER. Why? 99% of CPU sales are in the server/work station market ( which doesn't OverClock ) or the average consumer market which doesn't know about or doesn't care about OverClocking or are just too scaered to.

The only people that will really see the TWKR CPU's and understand thier purpose are here and places like this and tbh it kinda dissapoints me to hear about a CPU that unlike its hand picked memory and VGA cousins will never be within my grasp.

Memory companies do this but they release thier hand picked parts to the public like ATI did with those pimpin highly OC'd cards in the sexy brief cases a few yrs back.

AMD = pissin me off with this publicity stunt.
Click to expand...

Who's saying 5ghz+? These are all getting 6ghz+ on 4 cores. Seriously, how many people do you know personally have the knowledge and the gear to run LN2? The retail 955's out there are +- 6ghz on LN2.
 
Looks like another Ford vs Chevy deal.............the AMD guys love it, the intel guys think it is a joke.

Also it is like the racing dudes telling the public that their racing machine are just "slightly modified" street machine.

For the "science" of it I think it is cool, but means absolutly nothing for me in the "real world" as I really don't think I'd every be able to afford to have a set up like that in my home. :rolleyes:
 
jedihobbit said:
Looks like another Ford vs Chevy deal.............the AMD guys love it, the intel guys think it is a joke.

Also it is like the racing dudes telling the public that their racing machine are just "slightly modified" street machine.

For the "science" of it I think it is cool, but means absolutly nothing for me in the "real world" as I really don't think I'd every be able to afford to have a set up like that in my home. :rolleyes:
Click to expand...


Good way to reword my frustrations :) It is exactly the same as these companies coming out with these pimed out cars which have only the intent of showing what they "can" do although they know from the start they never will or by the time they do it will already be too late so in the end who cares? Its a tease, like some hot chick coming up to you and rubbing on you then walking away. Yeah she showed you what she "can" do but she refuses to do it leaving you to hump out your frustrations on the cat when you get home.

And honestly, I would rather have AMD vs Intel as AMD charges me less for the speed and being a broke SOAB is why my current rig is a PII 720 machine unlocked to quad. So I'm an AMD guy and I think its a joke because it wastes my time.
 
Probably a test run of AMD's binning proceedures.

They may be attempting to find a way to bin for the high end chips, so that way they can release 'extreme edition' versions of there chips at some point in the future.

This may be nothing more then exposure to build up a brand name.
 
Hard and Confused said:
All the more reason it shouldn't even exist. OverClockers in general make up less than 10% of PC users worldwide. Extreme OverClockers make up maybe .5% of that already less than 10%. Why even waste the time and resources marketing a product to .05% of all PC users worldwide?

I'm sure AMD did it to be able to go "Look what we can do" meanwhile Intel is laughing going "AND ITS STILL SLOWER THAN OURS ROTFL".

Worry about shrinking the clock for clock defecit you have against Intel's C2Q series first ( then the defecit you have on the i7's ) then you can worry about uber clocking CPU's AMD.

Now, if you'll excuse me I'ma go make a Michael Jackson love doll aimed towards kids. Because you know how theres such a big percentage interested in that and how its not waste of time and resources at all.
Click to expand...

What can an Intel CPU do better that an AMD CPU can't do?
 
alg7_munif said:
What can an Intel CPU do better that an AMD CPU can't do?
Click to expand...

Forget gaming and look at business apps then repeat the question to yourself. As I said, AMD has more important things to focus on rather than hyping a "might be released to the average consumer" CPU like increasing the performance per mhz of thier CPU's to at least match the Core 2 series.

Personally I use archiving programs a lot to send people files n such and though the $140 I paid for my PII 720 ( which unlocked ) is nice for the huge ass boost I got in Winrar archiving times but @ 2.8ghz per core I should at least be able to archive as fast as a 2.6ghz Intel Core 2 Quad but I can't. Given the Intel CPU's cost more but thats because they can as they are faster.
 
Hard and Confused said:
Forget gaming and look at business apps then repeat the question to yourself. As I said, AMD has more important things to focus on rather than hyping a "might be released to the average consumer" CPU like increasing the performance per mhz of thier CPU's to at least match the Core 2 series.

Personally I use archiving programs a lot to send people files n such and though the $140 I paid for my PII 720 ( which unlocked ) is nice for the huge ass boost I got in Winrar archiving times but @ 2.8ghz per core I should at least be able to archive as fast as a 2.6ghz Intel Core 2 Quad but I can't. Given the Intel CPU's cost more but thats because they can as they are faster.
Click to expand...

You can't?
http://enthusiast.hardocp.com/article.html?art=MTYyMSw0LCxoZW50aHVzaWFzdA==
 
quite a bit of trolling going on in here.

I especially like hard and confused acting like AMD speed binning some special chips is somehow holding them back from the work on performance they are undoubtedly doing.

like they should snap their fingers and be faster than intel, comes off as some whiney BS to me.
 
Someone in this AMD thread perhaps needs to consider funding a Mac system rather complain about what the marketing department of AMD is doing.

Hey, neat, the subliminal thing is a like Mad Magazine fold-in.
 
You must log in or register to reply here.
Back
Top