Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Buckwheat who wrote (41500)	11/14/1998 11:27:00 PM
From: Elmer	Read Replies (1) \| Respond to of 1573876

Re: "Could someone please explain how both AMD and Intel grade chips/wafers for speed bins. Also, if a whole wafer is graded for a particular speed, what happens to the chips that do not test out to the speed grade for the wafer? I might be under the mis-assumption that wafers are sampled to set speeds rather than testing of individual chips." Yes, I can explain. Die cannot be reliably graded for speed on a wafer. Speed binning must be done at final test in a packaged part. There are 2 reasons for this. 1) The electrical contact at wafer sort between the probe contact and the die pad is not as good as once the die is bonded to the bond plate in the package and 2) The worst case conditions exist at elevated temperatures which cannot be reliable reproduced at wafer sort because of thermal expansion and mechanical alignment. Some companies gather speed data at wafer sort but this is for statistical reasons only and no one grades an extire wafer at wafer sort much less an individual die, at least not the top companies. Some low cost suppliers (read ASIC houses) may try and do this but their quality levels are very poor. I do not believe AMD compromises quality in this area. At final test, packaged parts are tested at full frequency. Using a 450mhz part as an example, a part is first tested for opens or shorts in the package. Any failure would be the result of an assembly defect. After that, in varying order the device would be tested for fault coverage. I could write a paper on this one but suffice it to say this is the main test which screens for defects in the fabrication process. After that the part would be tested for functionality at speed. A 450mhz part must show full functionality at 450mhz, at elevated temperatures, and at minimum and maximum voltage levels defined in the data sheet. Should it fail, it would then be retested at 400mhz under the same conditions on down to the minimum acceptable frequency. If it still fails it's shit canned. Assuming it passes, and the vast majority do, there are more tests for drive capability, voltage level thresholds, buffer leakage, current draw etc. If the part passes the entire flow, it is binned out at the maximum frequency it operated at. 450mhz, 400, whatever. Some companies do not test at full frequencies but instead rely on indirect methods to guarantee operation at speed. I do not believe AMD compromises in this area either, however in regards to the K7, it will be interesting to see how they test the BSB at > 500mhz when that capability does not exist in a production environment. At any price. EP