Intel Corp (INTC)

[Unkwn]

Only transistor density matters.

Ok, got it. So you're finally admitting that your "competition only shrinks in one dimension" was made up BS from you and you finally read some actual facts about the latest process generation. That's good, really!

Now, regarding transistor density: There really is not such a thing. There's a big difference between logic and SRAM, as well as speed and voltages those transistors run at. Such have a big influence on how to read those numbers. Things you don't know but should know when you try to compare nodes.

If anything, there is logic density. Again, this is not so simple to compare since it is significantly influenced by design aspects, like speed, voltage etc. Layout and the tools play a big role. In the past, wiring wasn't so crucial for shrinking circuits. Today it is. Doesn't help much if you can shrink your transistors but can't connect them using the same area. Bringing in layers on top only goes so far, since you need vias and routing gets more complex and additional metal layers increase costs.

A much better comparison is SRAM density, since those are the most packed circuits on chips. They are tuned heavily towards density to reduce area as much as possible. Now, here, as you have seen in the article you posted yourself, TSMC and Samsung actually get a better density at 7nm than Intel at 10nm, which easily tells you that they don't have larger transistors than Intel, not in any dimension, as you were trying to spread before as your "fake news". By the way: I thought you were no Trumpian, so why do you use his language to try to deceive people in the same way? Certainly not working with me!

From the numbers in the very same link of yours, you can easily see that the differences between Intel 10nm and TSMC/Samsung 7nm pretty much are a wash. One is more dense there, the other is more dense here. Pretty much the same, at least when those numbers translate into similar designs, which is still to be seen once they are out.

7nm is not yet shipping in CPU's.
Intel's 10nm is shipping in mobile Cannonlake CPU's.

That is true, but this is with low frequency, low power small area designs where Intel obviously fights with the defects. In your post before, you made it seem the problem would only affect graphics. You wanna know why graphics are affected? Because the defects occur typically equally distributed in the design and graphics are typically consuming, together with SRAMs, the largest part of the on-chip area. With SRAMs, you can just disable parts of it and sell it as a lower performance part (something that Intel does heavily with its Celeron chips for instance). Even if one of the processor cores is affected, you turn that off. But with graphics, it is more difficult. That one is either working or not, increasing likelihood of it being affected by a high defect/non mature process. That's exactly what Intel is having problems with at the moment.

Frequency is the other issue. If you don't have full control over your new process, you won't be able to clock it to its highest frequency, since you're not in full control of your process/timing corners.

Intel admitted all this and already stated, that they will have it fixed by end of 2019. That's when they will be able to release high performance parts, like server chips, on its 10nm process.

Now, where's the problem? Besides Intel having promised to release this process already many years ago (I think it was initially meant to be a 2015 thing), the problem is that AMD announced a full performance, large area server chip on 7nm for beginning of next year (sampling by end of this year). Now you could simply state that they are overpromising. The thing is, not them would be overpromising but their foundry partner (TSMC in this case). According to AMD, the design has already taped out. They have contracts, you know, and it would cost TSMC a lot of money and image if they wouldn't keep their promise. That's why TSMC pretty much did stick to its timelines and delivered in the past. They really don't have another choice. A big customer like Apple would otherwise mean a lot of trouble to them, if they couldn't. So chances are pretty high they will stick to their schedule and this means not just some trial and error crippled parts that are barely usable like the ones shipping from Intel, but real performance large parts. Totally different league. Intel basically is selling its leftovers of the process trials that are somewhat usable in the open market. That's it, nothing more.

Therefore, these numbers for TSMC and Samsung at 7nm are
comparable to Intel at 10nm transistor density. In fact
the TSMC and SS ARE just 10nm. Claiming its 7nm and not
10nm is a lie.

You're not listening. Nobody ever denied that here on this board. I always stated that Intel 10nm and TSMC/Samsung/Glofo 7nm are pretty much a wash. Intel lost its lead just because they will be in volume production at a similar time as the foundries. A big difference to the past, where Intel was one or even two nodes ahead of the pack. They don't have that benefit anymore, which makes it much more difficult for them to bully AMD out of the market like they did it more than 10 years ago. The biggest problem for Intel is costs. It costs ever more money to develop those leading edge processes and Intel has no increasing volume to finance it. Since Intel missed its opportunity to make good use of those extremely expensive fabs (approaching 10 billion each), for volume businesses like mobile, foundry, memory etc., they now have an increasingly big cost disadvantage to the fabs. The fabs can benefit from all those growth businesses to pay for the development of their new fabs. Samsung even has a huge memory and in-house business together with its growing foundry segment. That's a completely different level of scale than Intel has, just because Intel screwed it up big time in the past years.

Now, AMD can just piggyback on it and concentrate on its designs, while leaving the costly process buildup to the bigger players. That makes them suddenly a vital opponent to Intel, which can (and does) even compete in pricing. Tough times ahead for Intel, very tough times. Sleeping giant.

The moderators only allow me one post per day. That's great
as it moderates me from wasting time interacting with clowns such as yourself. You should go read some basic math books, and get your
nose out of Investors Business Daily.

Hopefully you take that day off to try to actually digest the information I have given to you now and stop making up stories about foundries only shrinking in one dimension. Your sophisticated math skills obviously didn't prevent you from doing it.