News Intel Preps Software Defined Xeon CPUs: Buy Now, Add Features Later

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
AMD also doesn't do stupid segmentation, like this. The earlier nickel and dime comment is pretty accurate. Especially in the Covid era, companies are more cautious with their money, than in years past. Intel thinks it can still get away with pricing, that they used to do, when AMD was down. Even if AMD was just as fast, as Intel, the AMD chip would probably still be chosen, as you get all the features, for less than what Intel will provide. Intel's problem isn't necessarily performance, it is price/performance related. This is something Intel does, across most of its product stack. They have gotten better, at least, in the enthusiast space, as they were forced to reduce pricing.
AMD does segment in their server products. You pay significantly more for 2P versions of Epyc CPU's than you do for otherwise identical 1P versions. AMD doesn't have any manufacturing capabilities of their own and what they do get from TSMC is magnitudes lower volume than Intel is producing. That's why they don't have 9 million sku's like Intel does, they lack the manufacturing flexibility and the volume. AMD wants to make money like Intel does, not be your friend. It's simply ridiculous how some of you put AMD on a pedestal like they have some altruistic company guidelines despite acting just like Intel and Nvidia when they gain the advantage over those two.
 
I am not putting AMD on a pedestal. They have been more consumer friendly, with regards to price/features, than Intel has. If the roles were reversed, I would say the same about Intel. I have both AMD and Intel machines, as well. I just pick what is best price/performance at the time. Up until my 3700x, that I bought on the week of release, I had been using Intel, since the core 2 days. I didn't like Intel's business practices, but performance mattered more.
 
Yeah, but doesn't this go both ways?!
If I have a business where the base model is the exact same as the one with all the DLC because none of my software needs any of the extras, now I get to pay much less for the exact same performance.
No?

It's not as simple as you make it out to be, but if you want all of the cache of the CPU and just don't want AVX512 shenanigans, then you could do it now, I guess? Assuming they'll be differentiating the segment like that? I guess it's cheaper to make a "fit all cases" CPU than different lines that are more specialized. It solves many headaches at the expense of the per-niche performance, if that makes sense? Although, I'm pretty sure Intel is not stupid enough to neglect the big user base (as in, use case) for their Xeons.

Regards.
 
Difference is: there's always been an understanding that for the lesser products, it is because they didn't make the cut to be the fully enabled die. While you could argue that is not such the case anymore, it is still a hard pill to swallow for many I'd say? Regardless, DLC for CPUs: you get the full thing behind a paywall instead of the "best you could get" right off the bat.

Regards.

Maybe your understanding. The reality has always been that Intel can produce far more top-tier silicon than there is a demand for. So they take chips that would otherwise be the big-money parts and cut them down to a tier that needs more volume. Every silicon production facility does the same thing. Yeah, defective parts can also be re-binned down to a lesser SKU but that still doesn't give enough volume.
 
Maybe your understanding. The reality has always been that Intel can produce far more top-tier silicon than there is a demand for. So they take chips that would otherwise be the big-money parts and cut them down to a tier that needs more volume. Every silicon production facility does the same thing. Yeah, defective parts can also be re-binned down to a lesser SKU but that still doesn't give enough volume.
No, it is not "my" understanding. That's how it's been for years and years. As of late, with a very mature 14nm, I'm sure they have had very good yields, but for newer nodes it is always the case.

Regards.