Dear Intel: about the performance slowdown due to Meltdown/Spectre

[ppara5]

I was under the impression that Intel produced processors as follows. It tested all chips and ranked them according to quality. The top performers became i7-xxxxK. The next best became i5-xxxxK. The next best became i3-xxxx. And so on. So a chip of Pentium-quality could not be made into an i5, but the silicon might be good enough to upgrade into a lesser i3. And I'm assuming that it is firmware / microcode that separates i7 from i5 from i3 from Pentium from Celeron.

Given the above, here's one way to partially sidestep the litigation tsunami coming your way. Offer an upgrade path for those of us with older processors, possibly limited to Core. As an example, perhaps an i5-2405S could be converted into an i5-2400 but retaining HD 3000 graphics. We would ship our processor to Intel, the firmware / microcode would be modified, and the processor would be returned. Intel would need to supply prepaid shipping labels too.

[AlHill]

Sorry, but this is a ridiculous discussion.

Doc

[ppara5]

Ridiculous for which reason?

1) Processors cannot have their firmware / microcode changed after some event before Intel packages them for sale, i.e. they have been finalized.

2) Intel does not want to do it because it would cost a great deal of money.

If # 1 is true, then I completely understand, as it is a technical bar. But if # 2 is true, I disagree, as Intel could tell the many lawyers drooling at the prospect of possibly the largest class action suit in technology that it is mitigating the problem. Remember that Meltdown affects almost all Intel processors, along with a few ARM ones, but no AMD processors, so Intel's argument that it affects all vendors is simply not true. The two varieties of Spectre do affect all vendors, however.

My proposed solution would not be perfect or fair, as Linus Torvalds stated that performance degradation will be 5% for many people, but applications that perform many system calls, e.g. database servers, will suffer much worse.

[DChil2]

What you are probably misconstruing is the practice of binning which is the process of testing each die to figure out what speed it is able to run at.

Additionally you also need to understand the concept of Thermal Design Power (TDP). An i5-2405s likely has the same die as an i5-2400, but the TDP of both is very different.