When did CPU steppings become irrelevant?

[Sheninat0r]

I saw the post about the old Q6600 G0 SLACR and was reminded about my own i7-920 D0 SLBEJ. I feel like back in the day there were somewhat significant differences between steppings on processors, to the point where people advised me (and I agreed) to wait to build my computer until the old C0 steppings were all sold and buy when D0 was widely available. Thinking back to the last few processor generations, this hasn't happened; is it because Intel stopped making mistakes on their first production runs, or what?

 

[RaistlinZ]

I don't know, but I like it better this way. I hated having to try and hunt down a "good" stepping ever time I went CPU shopping.

 

[ninaholic37]

I think they call them "refreshes" now. Bay Trail has gone through quite a few.

 

[MongGrel]

I still use my old I-7 920 D0 SLBEJ for the HTPC in the bedroom, I've been too lazy to rip it out and put the L5690 in it

My sister-in-law still uses my old Q6600 G0 SLACR.

 

[YBS1]

Possibly the lack of competition from AMD at the high end enabled Intel to take more time coming to market eliminating the need for stepping improvements??? *shrug*

 

[Roland00Address]

Intel tick tock.

The philosophy of tick and tock roughly started in the 2006 / 2007 timeframe (architecture change followed by process changed followed by architecture change)

Q6600 was released Nov 06, right when they were just starting tick tock.

Steppings still happen but they are less important for either you know the process really well or the CPU design really well. Thus steppings are more important for internal knowledge than outside benefit. Furthermore you can kinda hide steppings in refreshes for example the new haswell refresh. We got lower tdp processors in tablets and ultrabooks and better overclockers.

 

[biostud]

Intel i7-920 D0 would be the last, I think.

 

[Idontcare]

I saw the post about the old Q6600 G0 SLACR and was reminded about my own i7-920 D0 SLBEJ. I feel like back in the day there were somewhat significant differences between steppings on processors, to the point where people advised me (and I agreed) to wait to build my computer until the old C0 steppings were all sold and buy when D0 was widely available. Thinking back to the last few processor generations, this hasn't happened; is it because Intel stopped making mistakes on their first production runs, or what?

To be sure the vetting process that ICs undergo nowadays is vastly more superior and advanced compared to the validation processes that were in place nearly 10yrs ago (yes folks, as scary as it sounds 2006 and the first quadcore are rapidly coming up on their 10th anniversary , do you feel old yet? )

Steppings are ridiculously costly. Both in salary compensation for the design team that you would be assigning the job of iterating the existing stepping, as well as in terms of absolute expenses of creating the new masks themselves.

So the ROI for creating a new stepping has really got to be there before management can justify the associated costs in today's business environment.

 

[tential]

I don't know, but I like it better this way. I hated having to try and hunt down a "good" stepping ever time I went CPU shopping.

Same, extremely happy it's not a huge thing anymore.

 

[Homeles]

To be sure the vetting process that ICs undergo nowadays is vastly more superior and advanced compared to the validation processes that were in place nearly 10yrs ago (yes folks, as scary as it sounds 2006 and the first quadcore are rapidly coming up on their 10th anniversary , do you feel old yet? )

Steppings are ridiculously costly. Both in salary compensation for the design team that you would be assigning the job of iterating the existing stepping, as well as in terms of absolute expenses of creating the new masks themselves.

So the ROI for creating a new stepping has really got to be there before management can justify the associated costs in today's business environment.

It's certainly harder to justify today, given that the bleeding edge processes use double patterning so extensively.

 

[Deders]

I read on here somewhere that later steppings of devil's canyon overclock better, and someone paid extra for the shop clerk to search through the boxes until he found one.

 

[Grooveriding]

I read on here somewhere that later steppings of devil's canyon overclock better, and someone paid extra for the shop clerk to search through the boxes until he found one.

That is not a stepping though, but likely rather looking for a chip made recently by checking the batch numbers. Chips tend to improve over time as they refine the existing process is the impression I've gotten. The last stepping I remember is i7 920 D0 and the D0 was a lot better than the C0.

 

[dmens]

Specifically in regard to the i7 920, there was a *lot* of effort on D0 to improve bin splits from C0, which increased the chance of getting a "good chip" compared to C0.

Fun memory: After the first overclocking demo for Bloomfield D0 at some IDF (I forget which year) I managed to snag the cherry-picked chip used on stage and it was in my home system for many years until it was replaced with a Haswell last year. I ran it 4.2ghz at 1.18V on a Noctua heatsink but it went much higher at the IDF.

 

[Mand]

Can someone explain what a stepping is and why it was done?

 

[Phynaz]

Mand,
http://en.wikipedia.org/wiki/Stepping_level

 

[Techhog]

Intel tick tock.

The philosophy of tick and tock roughly started in the 2006 / 2007 timeframe (architecture change followed by process changed followed by architecture change)

Q6600 was released Nov 06, right when they were just starting tick tock.

Steppings still happen but they are less important for either you know the process really well or the CPU design really well. Thus steppings are more important for internal knowledge than outside benefit. Furthermore you can kinda hide steppings in refreshes for example the new haswell refresh. We got lower tdp processors in tablets and ultrabooks and better overclockers.

The interesting thing is that Haswell Refresh seems to be the same stepping.

 

[crashtech]

Afaik, all Haswells are the same stepping. The "Refresh" is packaging tweaks and higher bins.

 

[ElFenix]

we probably also overblew the importance of particular steppings back in the day as well.

 

[Homeles]

we probably also overblew the importance of particular steppings back in the day as well.

Hey, free performance. That's part of what overclocking's all about.

 

[ElFenix]

Hey, free performance. That's part of what overclocking's all about.

it's not free when you're spending a bunch of time hunting down a particular stepping code and often paying more because it's in demand, much of which is based upon internet rumor.

 

[Tsavo]

All the low-hanging fruit have long been picked, eaten and shat. All that's left are small gains for giant dollars.

 

[zir_blazer]

What everyone in this Thread calls Stepping, I call a Revision. I associate Stepping with the five letter code that AMD Processors had which you could only know by looking directly onto the Heatspreader. There was a lot of mystique surrounding those because some Stepping codes were usually know as very good overclockers compared to others. It also allowed you to identify the physical die than a K10.5 had to know beforehand if there was any Core/Cache L3 that you could unlock.