131–140 of 160 posts
Earlier quoted context omitted.
Probably correct but making mobile CPUs is still a level below making the monsters our desktops are blessed with(Think of all the different SIMD etc. extensions the average Intel CPU) has these days
Virtualization support as well.
Earlier quoted context omitted.
Games!? No Man's Sky had trouble with SSE 4.1 being a requirement, and ended up patching it out... let alone AVX, AVX2, or AVX512. Typically, games won't take advantage of new instruction sets until it's ready to be a minimum requirement, as otherwise you need to maintain two code paths, the benefit of which is only to be speeding up execution on what are already the more powerful CPUs. What games can benefit from AV…
Some LLVM compilers produce code that (possible with GCC too) can jit itself on a per function on first execution e.g. -march=native so no branches required. This requires good planning of the game, i.e. this reduces inlining so one needs to be careful with putting the important bits in the JITted code
FreeBSD kernel uses ifuncs for SMAP, XSAVE, ERMS (Enhanced REP MOVSB/STOSB), RDSEED.
Earlier quoted context omitted.
Some LLVM compilers produce code that (possible with GCC too) can jit itself on a per function on first execution e.g. -march=native so no branches required. This requires good planning of the game, i.e. this reduces inlining so one needs to be careful with putting the important bits in the JITted code
If you're referring to ifuncs, they aren't doing anything like "JIT itself": your rtld simply calls your function to decide which version of the function you want to call and updates the PLT. FreeBSD kernel uses ifuncs for SMAP, XSAVE, ERMS (Enhanced REP MOVSB/STOSB), RDSEED.
Edit: I checked and they're not ifuncs
Earlier quoted context omitted.
Enterprise and data centers must consider the risk of a supplier not been able to fulfill their contracts. AMD is inherently more risky in the long term than Intel as AMD does not produce chips themselves. So they must offer cheaper products just to account for that. Add extreme inertia in the enterprise sector and one can see why AMD cannot have the same profit margins as Intel.
Both Intel and GF had problems ramping up their new node. Intel has been stuck with chip shortages for the past year and is still on 14nm because of this, AMD was able to (partially) switch to TSMC. I'd say that AMD's ability to pick a vendor and potentially multi-source for fabrication puts them at significantly less risk of supply shortages.
Earlier quoted context omitted.
Note, that the list price for Intel Xeons is something that nobody really pays. Behind the closed doors bulk pricing is negotiated and the price-per-chip in practice is only a fraction of the list price.
Enterprise sales: everyone gets 40% off list. Still, 60% is a big multiple of these prices.
There is no oxygen down at 8k for Intel, they will have to move to chiplets to get the yields up, to drop the prices that far. AMDs costs for these parts is probably double what the consumer versions are, so we are probably looking at 7k+ of profit per chip.
They are forcing Intel to move to chiplets or be really really wounded. Clearly AMD is not participating in a duopoly game, which would be the expected anticompetative behavior.
If I were AMD I would give away the mobos. This is going cause a lot of Intel parts to be EOLd early.
I'm curious why they aren't charging more for it. Wouldn't that be free profit margin? Or perhaps they are trying to hurt Intel by selling it at a price where Intel can't make a profit if they lower their prices to be competitive? Or, maybe it has to do with trying to quickly grab market share. Maybe AWS and other purchases of servers have a somewhat fixed budget, so cheaper chips translates directly to more chips so…
Intel is the standard and that has amazing sticking power. AMD is trying to make an offer DC operators can't refuse to break that. Intel is reeling from major design flaws and process stagnation right now. It makes sense for AMD to punch hard. Now is the time. Also note that the eternally predicted ARM64 wave into servers, workstations, and cloud is not materializing. So far nobody has been willing to build such high…
I don't believe 64 core, 3.35GHz at 200 Watt. Previous numbers for 64 core were around 2.2GHz. Looks like clickbait to me.
Earlier quoted context omitted.
> a new ISA (for server workloads) is a much harder sell than a new chip brand It really depends. As long as performance is there and the apps work as expected, I'll deploy my workloads on whatever runs them for the best cost per transaction. Additionally, we don't have a say on what architecture our hosted services are on - if AWS decides my RDS databases are to move to ARM64, or Google decides my CloudSQL will run…
You are in the minority. Most IT and devops people are massively change averse and tend to blame all problems on the most recent change. They would rather pay more for a known quantity. This is even more the case for non technical management.
Earlier quoted context omitted.
Black market, not grey market. QS are loaned on terms that explicitly prohibit resale. If you buy them, you are purchasing stolen goods.
No, you are not. You voiding your contract with Intel does not make me a criminal. At best this is a civil tort.
Trafficking in stolen property is illegal in all 50 states, although the severity and specifics vary. If you buy them across state lines, 18 U.S. Code § 2315 also applies, making it a federal crime.
And taking and selling the QS/ES chips is not "voiding your contract". It is theft, and given the value of the chips, would qualify as a felony.