"Intel Confidential" CPU purchased

[JFAMD]

That sounds fishy. Dell cannot legally sell ES chips in their systems. There are pre-production systems that are evals and will have ES chips, but not production systems. There are plenty of processes in place to keep this from happening, I spent a long time in their product group.

 

[punker]

Yeah..you dont know what the heck you are talking about....

I received one from a person who reviews chips and obviously iNtel wouldn't send him a disfunctional chip when they want a decent cpu review....

I received it when quadcores were not even available to get. So ofcourse it was a c0 stepping Kentsfield. However this was a cherry picked chip because I was able to hit 3.5ghz without going over 10% vcore boost. At the time most were having trouble getting their Q6700 and QX6700s much past 3.2ghz...

So you said they are neutered .... wrong
So you said they are terrible ocers .... wrong

Have anything else you can pull out your arse?

PS: that chip is still running today adn still overclocked. I have had it on 4 different motherboards. I received it late 2006.

engineering samples are prototype cpus

http://www.techarp.com/showarticle.aspx?artno=407&pgno=2

 

[Duvie]

engineering samples are prototype cpus

http://www.techarp.com/showarticle.aspx?artno=407&pgno=2

actually like in my chips case and the fact it was a c0 stepping which was the stepping of the first chips released in retail was beyond prototype...

Like I said it was sent to a site who reviews the chips (just before they hit retail channels). I dont think they would send him anything less then a final production chip.

 

[finbarqs]

ahhh brings back memories.... ES are usually free to the reseller's, and sometimes, they can be SUPERIOR to the production samples!

 

[evolucion8]

actually like in my chips case and the fact it was a c0 stepping which was the stepping of the first chips released in retail was beyond prototype...

Like I said it was sent to a site who reviews the chips (just before they hit retail channels). I dont think they would send him anything less then a final production chip.

Yeah..you dont know what the heck you are talking about....

I received one from a person who reviews chips and obviously iNtel wouldn't send him a disfunctional chip when they want a decent cpu review....

I received it when quadcores were not even available to get. So ofcourse it was a c0 stepping Kentsfield. However this was a cherry picked chip because I was able to hit 3.5ghz without going over 10% vcore boost. At the time most were having trouble getting their Q6700 and QX6700s much past 3.2ghz...

So you said they are neutered .... wrong
So you said they are terrible ocers .... wrong

Have anything else you can pull out your arse?

PS: that chip is still running today adn still overclocked. I have had it on 4 different motherboards. I received it late 2006.

I checked in the Intel spec site and didn't find any Quad Core using the c0 stepping in the 65nm manufacturing process. It was only used in the Penryn based QX9775 and QX9650 which are slightly inferior in terms of overclocking compared to the C1 stepping like the QX9770 and QX9650 v2. The best stepping in the Core 2 Quad 65nm was the G0, which was used originally with the QX6850 (Which I owned and went up to 3.60GHz and I was severely cooling limited), and the second revision of the Q6600 which overclocks the same. The B3 stepping was horrible in terms of overclocking, the same goes to the C0 stepping of the Quad Core based Penryn, the best stepping was the E0 which was used on the Q9550 and the Q9650, uses considerably less power and overclocks better than any C0/C1 penryn.

Reference:

http://processorfinder.intel.com/List.aspx?ParentRadio=All&ProcFam=2559&SearchKey=

http://processorfinder.intel.com/List.aspx?ParentRadio=All&ProcFam=2774&SearchKey=

 

[Gillbot]

Ahh, I love the ES speculation threads.......

 

[dmens]

Ahh, I love the ES speculation threads.......

People can just ask, there's more than one person in this forum who can provide correct info.

Just keep in mind that silicon variation for one stepping far exceeds the difference between different steppings, so getting a good overclocker is still a complete shot in the dark. The biggest average speed gain I have seen between 2 production steppings is 150Mhz, whereas top bin to bottom bin for your average product is maybe 1Ghz difference or thereabouts.

Another thing, people who buy the cheapest bin and plan to overclock will see the biggest speed variations because many of the cheapest chips are actually downbinned from the highest grade to satisfy demand, since not many people buy the $1000 chips. For example, i7 930 has some of the fastest chips, but plenty of junkers. You can tell by looking at the Vid, the junkers are like 0.2V higher.

IMHO it is a waste of time to hunt for specific steppings or product codes, afaik all binning info is completely internal. Best thing to do is check the Vid and compare against a known database.

 

[JFAMD]

Not necessarily sure on how much downclocking happens, it is very dependent on how you are running your fabs, what you are targeting, etc. combined with the overall market demand.

Let's say that you need to satisfy demand for a lot of high bin speed parts. You run your fab hot, your yields go down, and most likely you end up with parts that are netting out at their true speed.

If you are driving for volume, you will probably run your fab cold, get better yield but lower clocks.

In the first case, because of the fallout (and the fact that you are looking for higher clocks), you will probably have poor overclockers because each of the die is pushed closer to the limits (less headroom). This is regadless of the marked speed.

However, in the second, you could get more headroom because you might have more than your distribution of mid bins and some of that yield can be pushed lower.

But the to all of this is how many parts did you need to yield at what speed from the wafer?

The net of all of that is that you can never assume that you will get better overclocking from lower speeds. It has just as much to do with the recipe (input parameters) as it does with the distribution curves. It is probably true that you have better odds with lower speed, but you might not see as large of a proportional gain. Your mileage may (will) vary.

Processors are snowflakes, proceed accordingly.

 

[aigomorla]

*go hugz my NDA documents*

oh wait... those dont apply anymore..

As others said, ES's are all on loan from intel.
They are property of intel.

ES's are usually intend to give there support partners and vendors a test unit for inhouse.
Sometimes some companies at such inhouses will no longer require these processors so they give them to there employees once the NDA agreement is over.

There employee's will then try to sell the cpu.
This is not liked by intel.

1. Chances are if ur running an old ES, intel wont care. The cpu has been reversed engineered and most likely its full of patents.

2. If you had a gulftown, Intel will spank you. Gulftown's were intel's crown cpu, so they did NOT apreciate people showing them off pre launch.

3. If you plan on ebaying the cpu... Well Intel has been known to buy the cpu off EBAY to trace the serials and see which partner messed up.

Chance are, you will never see another cpu from the person who you bought it from again, unless he gets new sources.


ES cpu's are not special unless u got one prelaunch.
Then yes they are special because you got the "i have one you dont" factor.

If i had the option to get a retail over an ES i would. Personally i ask my source for production samples.... ES's but finalized ones which are meant for production.

Example..

The gulftown A0 stepping was NOT intended for consumers. This was a straight inhouse CPU, which was phased out by B0, and B1 steppings.

B0 & B1 is the 5600 series and 980X u see today.

My advice... unless the guy who u are getting the ES cpu from is a known and trusted source... dont bother.

If something goes down, and intel finds your source.. You will get a nice email from intel security asking for the cpu back without a refund. And you must comply because intel will file it as a stolen cpu.

JFAMD... can you tell us the offical book policy on AMD ES's tho.
Im assuming its the same as intels?

 

[JFAMD]

I don't know the policy but I know we take those things very seriously. I'm amazed at some of the things our company can find out. Security is a very important thing in this business.

 

[Danyune]

well i guess i should hope intel doesn't pursue

 

[JFAMD]

Pursue an individual who has 1 processor in their posession that paid for it? Probably not likely.

Pursue the individual who got the processor at no charge, agreed to keep it in their posession and then turned around and sold it for cash with 100% profit? Far more likely.

When you receive samples from us it is made clear that you are supposed to use them for their intended purpose only. You are not supposed to even give them away at no charge or let them leave your premises.

 

[ghost recon88]

well i guess i should hope intel doesn't pursue

Odds are good they won't. Just don't bring it up in conversation or mention it again :hmm:

 

[Gillbot]

ES cpu's are not special unless u got one prelaunch.
Then yes they are special because you got the "i have one you dont" factor.

That's what i've been trying to say for years yet everyone thinks they are special uber overclocking unlocked monsters. Sadly, this is far from the case and retail chips usually perform better though ES's seem to take more "abuse" without popping.

Pursue an individual who has 1 processor in their posession that paid for it? Probably not likely.

I've seen it happen. It really depends on the chip though. The farther out in retail it is, the less likely it is to happen though. If you got something pre-launch though, odds are intel WILL want to know where you got it from and they will pursue after ONE cpu. Search around, there are plenty of threads elsewhere and topics on the internet about that exact scenario.

 

[Suspicious-Teach8788]

Do you want to sell it to Gizmodo before the po-po knocks at your door? Pocket a quick $5,000 now!

 

[Shmee]

I am using a B1 980x ES at 4.3 GHz. Would it be wise to ask for a retail replacement and send it back?

That's what i've been trying to say for years yet everyone thinks they are special uber overclocking unlocked monsters. Sadly, this is far from the case and retail chips usually perform better though ES's seem to take more "abuse" without popping.


I've seen it happen. It really depends on the chip though. The farther out in retail it is, the less likely it is to happen though. If you got something pre-launch though, odds are intel WILL want to know where you got it from and they will pursue after ONE cpu. Search around, there are plenty of threads elsewhere and topics on the internet about that exact scenario.

I seem to understand from aigomorla and rubicon that the retail 980x's are actually much harder to pop than the ES's, especially the A0's. With that being said, would getting a retail chip increase OC potential? If so, then returning the ES for one is a no-brainer, as long as my contact can provide.

 

[JFAMD]

I am using a B1 980x ES at 4.3 GHz. Would it be wise to ask for a retail replacement and send it back?

Um, yeah, that probably isn't a good idea.

Dear Intel,

I have something that belongs to you and I have no right to have it. As a matter of fact, by having it I have denied you the opportunity to make the revenue that you should have made if I had bought a retail product.

I am sending it to you, please forego the revenue a second time and send me a free processor to replace it.

 

[aigomorla]

Would it be wise to ask for a retail replacement and send it back? .

I hate to say this with JFAMD watching.

But ... it doesnt hurt if they dont know.

And no your OC potential might be worse, and might be better.
Its still a lottery.

The B1 was intended for consumer sale.

 

[Gillbot]

I am using a B1 980x ES at 4.3 GHz. Would it be wise to ask for a retail replacement and send it back?


I seem to understand from aigomorla and rubicon that the retail 980x's are actually much harder to pop than the ES's, especially the A0's. With that being said, would getting a retail chip increase OC potential? If so, then returning the ES for one is a no-brainer, as long as my contact can provide.

From what people tell me, the A0's are preferred by some for benching because they are toughest but you need LN2 or heavy duty sub zero cooling. The later stepppings are better under "normal" air/water, etc. This is what i've heard from quite a few benchers but I have no first hand experience though.

As far as sending it back, if you have a legitimate reason for having it, there should be no concerns. Otherwise, your source will get into trouble, also as JFAMD says:

Dear Intel,

I have something that belongs to you and I have no right to have it. As a matter of fact, by having it I have denied you the opportunity to make the revenue that you should have made if I had bought a retail product.

I am sending it to you, please forego the revenue a second time and send me a free processor to replace it.

I hate to say this with JFAMD watching.

But ... it doesnt hurt if they dont know.

And no your OC potential might be worse, and might be better.
Its still a lottery.

The B1 was intended for consumer sale.

Exactly, you never know if it will be better or worse. Plus, if you keep your head low about the ES (shhhh, big intel is watching) they'll nover know you have it... Well, until now. You are on their radar, HIDE!

 

[JFAMD]

I hate to say this with JFAMD watching.

But ... it doesnt hurt if they dont know.

Every one of our processors has a serial number etched on it. We would know. We can trace everything back to its source.

You would never get a warranty replacement on a part that does not have a warranty, and engineering samples do not have a warranty.

 

[Duvie]

People can just ask, there's more than one person in this forum who can provide correct info.

Just keep in mind that silicon variation for one stepping far exceeds the difference between different steppings, so getting a good overclocker is still a complete shot in the dark. The biggest average speed gain I have seen between 2 production steppings is 150Mhz, whereas top bin to bottom bin for your average product is maybe 1Ghz difference or thereabouts.

Another thing, people who buy the cheapest bin and plan to overclock will see the biggest speed variations because many of the cheapest chips are actually downbinned from the highest grade to satisfy demand, since not many people buy the $1000 chips. For example, i7 930 has some of the fastest chips, but plenty of junkers. You can tell by looking at the Vid, the junkers are like 0.2V higher.

IMHO it is a waste of time to hunt for specific steppings or product codes, afaik all binning info is completely internal. Best thing to do is check the Vid and compare against a known database.

In the past I use to speculate on binning as the reason I purchased the lower p4 1.6a and clocked it to 2.71ghz...purchased the 2.4b and clocked it to 3.36ghz and purchased a 2.4c and clocked it to 3.5ghz....I even did this with my first A64 3000+ winchester that did 2.7ghz...and my first core 2 duo...

I figured basically intel doesn't build chips at certain speeds but one process to build that generation of chips....If the yields were good then they often had most of the chips capable of being the top speed in that line....however market demands and cpu line pricing would make this foolish for intel to only have the expensive chips....so they downclock the rest to fill in the supply of the slower chips needed at the given price points...

Ofcourse I was blasted by a few INtel regulars who said I was wrong (cough...PM...cough)...this doesn't happen...
basically everything you stated Dmens they said was BS

 

[JFAMD]

When I said "processors are like snowflakes" I was being very serious. Every processor die has minor differences in metals and layers that will make it perform ever so slightly different.

When you make wafers, every single die on the wafer is the same. But every die will come out with different characteristics that will yield a different performance (clock speed).

There are "recipes" for getting higher performance parts, but you get lower yield. You can target lower clock speeds and get better yeilds. Each wafer can have a different target. So, in essence, you are trying to maximize revenue, and you have 2 different levers to pull - quality and quantity (quality is not what you normally think..)

If you have a big demand for the highest speed SKUs, then you push the process limits, you get more of those parts, but you don't get a wide distribution nor do you get the best yeilds. Then you might have to run a few extra wafers "cold" just to fill out the demand that you have in other areas. The biggest part of this business is accurately capturing demand.

When we originally built Istanbul, the first 54,000 processors went to 2 deals, Jaguar and Kraken. They wanted top bin parts, so we had to run proportionately more wafers to satisfy that demand. That meant that we were tight on supply of top bin (because they took the majority of the parts) and the total number of parts yielded was lower than we expected because yield on "hot" wafters is not as high as "cold" wafers. But it was worth it and all ended up well.

You were correct that the design of the chips is the same. There is not a 2.4GHz design, nor a 2.4GHz "recipe". There is a single design and you can twist the process knobs in order to maximize your 2.4GHz yield.

 

[taltamir]

When you make wafers, every single die on the wafer is the same.

No it isn't, if that was the case there would be no such thing as die harvesting... when you make a wafer every single die is unique.

CPUs are manufactured via chemistry, you use lasers (shot through a mask to cause certain shapes) to affect the chemical reactions that cause the die to grow on a nearly atomic scale (we are talking one atom thick and a few dozen atoms wide here)... alternating between solvents and depositing materials.

It is very impressive, it also results in every single processor being "unique" (slight differences)

But every die will come out with different characteristics that will yield a different performance (clock speed).

huh? so are you saying they are the same or are you saying they are different?

 

[aigomorla]

No it isn't, if that was the case there would be no such thing as die harvesting...

ur joking right?

Each die is unique.

Which is why we tell everyone... your OC clocking on each cpu will be a new experience.

Have you ever pop'd in one cpu to see its identical to another?

In all the intel cpu's I've played with ... only 2 were close in being identical.
And they were from different batches.

CPUs are manufactured via chemistry, you use lasers (shot through a mask to cause certain shapes) to affect the chemical reactions that cause the die to grow on a nearly atomic scale (we are talking one atom thick and a few dozen atoms wide here)... alternating between solvents and depositing materials.

And as you are aware, yield issues come out when fabing.
The cpu's havested from the center were "known" to be the better oc chips from C2D which is why everyone wanted those low number steppings.

Steppings told the user which fab they came from, and approx which location on the waffer, which your chip was havested from.

If chips were all identical, then why did we go stepping crazy on c2d?

 

[taltamir]

When you make wafers, every single die on the wafer is the same.

No it isn't, if that was the case there would be no such thing as die harvesting...

ur joking right?

Each die is unique.

That is my point. Every single die is unique. The claim that "every die on the same wafer is the same" is false... although, my note about die harvesting was misplaced. Because if every single die on a wafer was the same you would still have differences on a wafer by wafer basis; thus the harvesting would be limited to lower quality "wafers", however that is all a "what if" thing, in reality every single die is different and it isn't a wafer by wafer thing...

My point was that every single die is unique, no two die are the same.

If chips were all identical, then why did we go stepping crazy on c2d?

Reread my post, I am clearly saying that chips ARE NOT identical. Each is different.

PS. I pointed out that JFAMD was contradicting himself, saying first that all the die are the same and then that they are all different. I think what he meant in the first part of the statement is that all the die on the wafer are the same design, but each one ends up being physically different.

I said that each die is different, you misread what what I said and told me that each die is different...
So each of us three agrees that each die is different, but we saw fit to "correct" each other based on misreading what the person before us wrote.