Everyone loves Intel engineering samples. They are special handpicked samples for reviewers that are more overclockable and have unlocked multipliers. Or do they? We take a look at the Core 2 Duo engineering samples and show you why they are not the processors that everyone thinks they are. Should you buy one, or not? Read on and find out. Link : Why You Should AVOID Core 2 Duo Engineering Samples
Very interesting article as usual indeed But I really get confused then. What efficient alternative do we have? I mean, how to "complain" or make things changes regarding the Intel politic?
Ah, I thought it was an old article, since I'm finding about these guides of eng. samples too. So are they sent to the reviewers for testing purposes only, or just a plain old prototype?
Looks like old prototypes. I wonder if it's anything to do with the recent warehouse hijack case in penang. No wonder Intel reported there's no impact to their C2D shipment schedule.
the ones being sold aren't really ES but merely marketed as ES. there's no way Intel releases so many ES out the industry to benchmark and test.
Well, you can't reject something that's not suppose to be sold in the first place Basically it's like what the article said, these are engineering batches and also very early batches of the chip. think beta testing.
Further clarification: Those sent out to reviewers during the pre-launch marketing hype were ES's tested to be 100% ok. Production variances allow 2-3% tolerances of defects or errors (Which explains why some processors can clock higher then others, and is actually why binning exists), which would under most normal circumstances, would not be encountered. These ES's were made sure so as not to cause any misrepresentation of the final product itself. Where else those flooding into the market currently is those taken off the production line randomly (I have no data as to how many per wafer batch as this remains P&C data) for Quality Assurance purposes and tested to the maximum severity of conditions. Some would be damaged beyond usage (in other words, dead) while some have survived but with some form of damage caused during testing unknown to us (Unless you have Intel's developer tools). According to my sources, these are usually loaned to Intel staff and are for them to use so as long as they are still being employed. There is also another source of these ES's which is through the validation and testing labs as these processors would be written off as consumables once tested, meaning is not of commercial value to Intel anymore. I have personally come across this (Prescott Pentium 4's, LGA775) which after testing for booting, the one's working were about 3 out of 10 (I had a tray of 20 to test). And those bootable even though multi unlocked, could not pass any stress test at speeds more then 5% of stock clocks [FSB x base multiplier (x14)]. The reason I didn't include this into the article itself as: 1.) No documented or substantiated proof (Info passed through word of mouth, and personal experience) 2.) What I said would be tantamount to libel 3.) No proof of whether such a practice is still being done And as for the hijack case, from what I've heard, was actually production units of desktop and mobile processors of Cedar Mill and Centrino variety. I may be wrong about that
Well written article. Seriously the Engineering Sample is for the boys in the Intel labs for testing purposes only. But not sure how it leaked out, maybe one of the boys gave it to someone else, or whatever it might happen.
This is a very good write-up. Anyway, AFAIK, ES units are not tested in validation lab. In fact, a randomly picked production units are tested as there is no point to test CPU which will not release to customers. ES units are simply some beta units for Intel employees to play around and feedback when they found some serious issue. These units are actually new and not stressed in lab before. It is just that the CPU is subject to some known serious logic bugs (such as 1+1=3 instead of 1+1=2) where production unit might have solve this problem. There is another comment on the C1E and speedstep: In Intel power management, there are a few sleep states in Core 2 Duo - C0, C1, C2, C3 and C4. (C2 to C4 supported in merom core). C0 is the performance state, whereas C1 is the HALT state. Tradisionally speedstep occurs in C0 state, where it dynamically changes the Multiplier and VID based on a pre-defined table inside the CPU. This is how the frequency and voltage changes according to CPU load. In Pentium 4, the C1 state only calls the HALT instruction to stop the execution units and its pipeline. However, registers, external bus logic and cache are still working at full clock speed and Vcore. In order to bring down the leakage current and idle power, C1E is introduced by combining C-states and Speed Step. So during C1E, the execution pipeline is halted and the clock speed + voltage are throttled to the lowest possible value (6x multiplier with lowest supported Vcore).
Thanks for the comments Yee Yong. That makes the C1e halt state easier to explain. Will add to the article later. As for the validation lab part, I may be wrong about which lab, but such a practice did exist but I don't think so anymore since Intel got pretty strict about production floor security.
You are welcome Eu Bian Anway, I've just finished reading the power management article in digit-life and it is kind of misleading. It described the sleep state and thermal state very bluntly (even some are incorrect) and i suspect the author didnt really understand the key concept of Intel power management architecture well enough. If you have any doubts, just PM me here/LYN or send me an offline msg in yahoo haha, is there a rating for your review? I wanted to give 5 stars
LOL, ratings don't really matter. As long it provokes people to start thinking, questioning, and contributing, I've already achieved my initial goal TBH. I know I suck at EE stuff (You know what major I'm doing eh Chargey?) so if it does cause some positive outcome and we can further improve on the article, great for the consumer IMHO
you are considered bloody darn good although u r not major in EE. So many major EE fellas can't even understand simple reviews in anand yeah, we consumer should be more informed on all kind of products. Online review and forum is a great place to share these info out.
Sorry guys if my statement earlier was unclear. Actually the Core 2 Duo i meant here includes Conroe, Allendale, Merom & Woodcrest. For non-mobile version, the CPU will have C0 and C1E state only, whereas, Merom (the mobile cpu) supports all the extra states such as C2, C3 and C4.
Why You Should AVOID Core 2 Duo Engineering Samples Rev. 1.1 The guide has just been updated! Everyone loves Intel engineering samples. They are special handpicked samples for reviewers that are more overclockable and have unlocked multipliers. Or do they? We take a look at the Core 2 Duo engineering samples and show you why they are not the processors that everyone thinks they are. Should you buy one, or not? Read on and find out. Here is the update:- Link : Why You Should AVOID Core 2 Duo Engineering Samples Rev. 1.1
