CPU manufacturing question

Oh man! Ad-blocking software has been detected! :'(

This website is run by the community, for the community... and it needs advertisements in order to keep running. Blocking our ads means your killing our stats!
Please disable your ad-block, or become a premium member to hide all advertisements and this notice.

I understand that when CPUs are made they are produce in mass and cut and then individually boxed. Is it possible for some of the CPUs to be slower/faster than what the rating says it is? Even if it is marginal?

 

Oh man! Ad-blocking software has been detected! :'(

This website is run by the community, for the community... and it needs advertisements in order to keep running. Blocking our ads means your killing our stats!
Please disable your ad-block, or become a premium member to hide all advertisements and this notice.

The wafer can have various defects, some of these can be worked around by turning off the affected units of the processor and selling it as a cheaper model. In the past for example you could get 486SX and DX processors where the maths co-processor was disabled. Now they can disable some instructions or an entire core. Most modern microprocessors have microcode instructions, these can be altered after the CPU is made using a flashing process, in this way the chip can be altered even after manufacture. I suggest reading up on chip yield and moores law.

 

Thanks for that.

I'll have a look into that.

 

See my old post here :-

http://www.certforums.co.uk/forums/thread18842.html#5

In particular this bit :-

http://arstechnica.com/articles/paedia/cpu/moore.ars/1

It explains how wafer defects, die size and gate density can effect chip yield.

The design of the CPU or microcode can also have bugs, or 'errata' as the manufacturers like to call them. The microcode can be fixed but the design CPU can only really be reconfigured one its made. Later versions have the bugs fixed and this is the stepping number.

http://hardware.slashdot.org/article.pl?sid=07/06/26/2152246&threshold=-1

 

Thanks a lot, I'll print that out and read it.

 

Yes, there is a process called 'chip binning'. Chips are made to a certain specification, say in terms of processor speed, and if they fail stability tests at that particular speed, they will be rated down to a speed at which they are stable and 'rated' at that speed at a cheaper price.

 

cheers, is it possible for them to come out faster?

 

See here :-

http://www.edn.com/index.asp?layout=blog&blog_id=1480000148&blog_post_id=1240018324

Yes some perform better some perform worse, theres probably a standard deviation. The manufacturing process is not perfect and will produce products that vary in quality.

http://www.cosc.brocku.ca/Offerings/3P92/seminars/260,2,Semiconductor Devices

Just bear in mind that the manufacturers generally know their products inside out, the better performing products are often branded and sold for a premium.

 

Yes they can come out faster, but only marginally. For instance with my Pentium 4 HT rated at 3.2GHz, it actually runs at something like 3.234GHz or something - so in essence when you are sold a chip you are actually sold its approximate clock speed. IIRC, it's near impossible for any fabrication process to get any number of chips to be rated at exactly the specification quoted.

Another reason why it is possible for them to come out faster is by the very nature that we can overclock our chips. Of course, the fabrication process and layout designs might not have specifically been designed for such clock speeds (hence the need for extra cooling solutions for extreme overclockers), but the fact that stability can be maintained suggests that improvements can be made to the chip design to improve clock speeds (shorter distances between gates, faster switching gates, silicon with thinner and smaller surface areas, etc.)