Yeah, the last such comparable situation for Intel was all the way back in the mid-2000s, when 90nm turned out to be a bust for producing anything that wasn't a Pentium M. And we didn't get any real indicator of how 65nm was going to perform until the first samples of Cedar Mill/Presler showed up and turned out to have reasonable power consumption and crazy-high overclocking headroom.
Until some actual 7nm samples show up, all we can do is guess.
What does fixing it by hand mean? In the literal sense, it's impossible to fix a defected chip by hand.
uzzi38
Platinum Member
- Oct 16, 2019
- 2,635
- 5,979
- 146
Yeah, I don't actually mean they like grabbed a random pin and started poking it, but I can't tell you what exactly they did. I don't know either lol, I just heard they put quite some time and effort into manually fixing a couple of dies.
I'm not trying to be difficult, but manually fixing a couple of dies just doesn't make any sense. I could see them painstakingly testing many, many dies to find a couple that were fully functional, that would make sense, but manually fixing a fabricated IC doesn't add up to me.
I'm not going to comment on the validity of this rumor. However, it is possible to "manually" fix a chip. I know little of the specifics, but I do know that it's possible trace back defects to individual transistors and "re-wire" the circuit to fix those defects. It's an extremely low throughput process, and very finicky, but it's possible.
Ajay
Lifer
- Jan 8, 2001
- 15,458
- 7,862
- 136
I don't think it's 'finicky'. Years ago, when my cube was located next to the ASIC dev guys, they would manually write the Verilog code. I don't recall their exact workflow, but I imagine it's something like this. Taking the existing cells and hand tuning tuning them to fix what ever bugs were present.
I'm not talking about writing Verilog, but rather taking an already-fabricated, defective circuit, and modifying individuals wires to work around or fix the defects.
Ajay
Lifer
- Jan 8, 2001
- 15,458
- 7,862
- 136
Seems weird. Base clock from 1.3 -> 2.8 on 15W and from 2.3 -> 3 at 28W?
Those numbers does not make sense on same node.
It's possible to deconstruct an IC layer by layer and then perform lithography etching to re-wire small portions of an IC. This has been done for many years to test high level metal fixes before doing a full redesign and refab. However, this is a design issue fix, not a yield or process issue fix, and you're not doing it on the low level metals, not at advanced nodes especially. If you can fix an IC through this process with consistent results, you can then fix the design and have a production ready chip. The only thing I can think of that would make any sense is if Intel messed up in their redundancy connections and were able to fix it on a couple of chips this way but then determined that even with redundancy, yields were too low.
Even without the comparison to Icelake it seems weird. 2.8 GHz base clock at 15 W and 3 GHz base clock at 28 W? I think it's more likely that 2.8 GHz is also at 28 W (or very near).
Edit: jpiniero beat me to it.
uzzi38
Platinum Member
- Oct 16, 2019
- 2,635
- 5,979
- 146
Probably just 25/28W.
Either way, it's an impressive uplift from just one generation. Actually means Tiger Lake should be pretty good.
Now just gotta wait and see what kind of clocks are sustainable in reasable power budgets.
Who told 28W for i7-1185G7? You would expect 8G7 and not 5G7 if it's 28W.
There is no confirmation, the naming scheme is more a hint towards 15W, it's a copy of 1065G7 Icelake. With you poor track record is also means 15W is more likely. Of course these models always could run on 25W cTDP up.
Intel's gotta be freaking out about Renior. So clocks (and power consumption) are going up, so they can say how great Tiger Lake is. 28, maybe 35 W at base.
The top 15W model would be a theoretical 1165G7.
Your postings are full of fanboy nonsense lately. Not long ago you wrote this in regards to SKUs:
It did not take long to get another example of you bright speculation skills. You are always wrong and I'm expecting this trend to be continuing.
Last edited:
Exactly. People complain about bureaucracy and efficiency when dealing with the goverment but it isn't any better in large organization. As we saw with AMD you can lay-off a large part of your employees and still deliver a competitive product. If you lay-off the correct people. Certain problems simply can't be solved by more man power (in fact that often makes is slower).
For laptops especially you have 2 yearly windows that you really want refreshed models available for. The first is back to school from late July to early September. The next is the holiday window from mid November or so through Christmas. Based only on that tweet, it seems like Tigerlake will officially launch in the middle of the year, but not in time to make the back to school window in terms of volume that OEMs want.
DrMrLordX
Lifer
- Apr 27, 2000
- 21,634
- 10,850
- 136
I was aware of that, but I didn't want to jump to the conclusion that Tiger Lake-U would be MIA until November or so. That seems kind of bad.
It also looks like their 10+ process used in TGL has better yields than the process used to make ICL chips. I think this bodes well for a transition to 10nm for the desktop in 2021!
It probably won't be MIA, but just in limited models until the holiday refresh. Again, just based off of that tweet.
TRENDING THREADS
-
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 10K
-
Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)- Started by Vattila
- Replies: 13K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 7K
Facebook
Twitter