Slashdot Mirror
Intel Cascade Lake-AP Xeon CPUs Embrace the Multi-Chip Module (techreport.com)
Ahead of the annual Supercomputing 2018 conference next week, Intel today announced part of its upcoming Cascade Lake strategy. From a report: The company teased plans for a new Xeon platform called Cascade Lake Advanced Performance, or Cascade Lake-AP, this morning ahead of the Supercomputing 2018 conference. This next-gen platform doubles the cores per socket from an Intel system by joining a number of Cascade Lake Xeon dies together on a single package with the blue team's Ultra Path Interconnect, or UPI. Intel will allow Cascade Lake-AP servers to employ up to two-socket (2S) topologies, for as many as 96 cores per server.
Intel chose to share two competitive performance numbers alongside the disclosure of Cascade Lake-AP. One of these is that a top-end Cascade Lake-AP system can put up 3.4x the Linpack throughput of a dual-socket AMD Epyc 7601 platform. This benchmark hits AMD where it hurts. The AVX-512 instruction set gives Intel CPUs a major leg up on the competition in high-performance computing applications where floating-point throughput is paramount. Intel used its own compilers to create binaries for this comparison, and that decision could create favorable Linpack performance results versus AMD CPUs, as well.
Intel chose to share two competitive performance numbers alongside the disclosure of Cascade Lake-AP. One of these is that a top-end Cascade Lake-AP system can put up 3.4x the Linpack throughput of a dual-socket AMD Epyc 7601 platform. This benchmark hits AMD where it hurts. The AVX-512 instruction set gives Intel CPUs a major leg up on the competition in high-performance computing applications where floating-point throughput is paramount. Intel used its own compilers to create binaries for this comparison, and that decision could create favorable Linpack performance results versus AMD CPUs, as well.
this first post lmao
Synthetic benchmark completely rigged to give Intel's kit an advantage does indeed give it an advantage, news at 11.
A 1.5 years ago: Glued together CPUs BAD
Now: Glued together CPUs GOOD
The AVX-512 instruction set gives Intel CPUs a major leg up on the competition in high-performance computing applications where floating-point throughput is paramount.
The nature of the AMD link is that is could add a co-processor to it's design easily. That's one of the reasons both Intel and AMD are going this way is because it makes it easier to adjust to the market.
Intel, circa 2017: "We cannot figure out how to successfully engineer 10nm wafers. Our tick-tock strategy is stalled, and we cannot design chips that are any faster. What should we do?"
Intel Solution: "MORE CORES!"
Intel, circa 2018: "AMD just released Ryzen, and it's destroying us in benchmarks. Anyone figure out that 10nm thingie yet?"
Intel Solution: "Nope. But we did add MORE CORES!"
Now must be a great time to be an Intel engineer.
But did they get around to fixing those horrible information leaks they designed right into their CPUs? If not, when will they get to it? Ever?
Meanwhile, it's easy to forget that IBM still produces faster CPUs with more cores and more (actual-SMT-)threads per core and more cache and more performance. Mere mortals can't actually buy them, but they do exist. intel has always been a poor man's game in comparison.
What are the implications for per-core or per-CPU licensing schemes? Its already bad enough as it is.
N/t
Intel says Intel CPUs are great. Yeah, what else are they going to say?
It is all marketing hype until independent third-party bench-marking is done.
Buy Intel!.
Fuck linpack. We need a new measurement standard. Vulnerabilities per release.
One of these is that a top-end Cascade Lake-AP system can put up 3.4x the Linpack throughput of a dual-socket AMD Epyc 7601 platform. This benchmark hits AMD where it hurts.
Now let's see what it costs.
how many pci-e lanes in 1 Socket and 2 socket?
With AMD you have 128 with one or 2
mac pro will have this late 2018 starting at 7-10K (dual cpu base with fully loaded ram channels and crap base video card)
Inmos linked over a thousand processors together. You could even have a Beowulf cluster of them!
Seriously, bung a large amount of RAM onto some chips, then add maybe 8 of these CPU chips, and put the whole lot in one case. Not quite System on a Chip, but it's close enough.
Specifically blocks non-Intel CPUs from getting an optimized code path, hardly shocking their CPU performs a lot better.
2018 Mac Pro?
this is a joke right.
intel is losing big time to amd
Does no one just ask - is this even a reasonable claim?
Intel is going to be on an OLDER process node - their architecture is not running 340% faster than AMD.
Intel is comparing their theoretical future chip with Epyc chips shipping now. https://www.newegg.com/Product...
Once these chips are available in quantity (they are not) drop them into some servers and start bench-marking them on performance per price /watt. And compare them to AMD chips coming out at that time.
If this is the benchmark that is hitting AMD where it "hurts" AMD is in good position. When your competitor benchmarks their future products against your current products instead of their current ones, you KNOW you are good.
And the latest set of benchmark shenanigans don't look good for intel either.
If the customer wants only 1-core CPU instead of 32 cores then does it divide 32 the price and the power consumption?
Why bother? I'm reliably informed from Slashdotters that you most people only read email and browse the web. Therefore these are unnecessary.
It’s all useless, until YOU, PERSONALLY have verified either their trustworthiness, or the thing itself.
Because YOU are who's watching the watchmen. (Or me in my case.)
And even then, you are not an independent entity, just like any third-party too. You are still subject to manipulation. Like using an Intel compiler (which is deliberately designed to make code slower on AMD CPUs), or leaving XMP disabled on AMD, or not watching out for built-in "benchmark mode" where the device cheats, etc.
Oh, and if you’re not an expert, or worse, *believe* you’re an expert when you’re not, you cannot even judge the competency of said third party.
Which is, of course, utterly nuts, ... especially for hardware that will, as always with Apple, be a total piece of shit on the inside (see Louis Rossman's videos), ... and hence probably actually true exactly for that reason.
To me it still sounds like a surreal joke.
I couldn't even get to such a low level of conscience if I ripped my brain out. I honestly thing we should define two separate species right now: Homo Sapiens, and Homo Psychopathis. Maybe Homo Iumentis too...
Whats the timeline for this? AMD already engineering samples of 64 core EPYC 7nm chips out there expected to come early next year. How long until the intel stuff is actually available, not just "teased plans"?
Straight from the article, "Intel didn't note whether Hyper-Threading would be available from Cascade Lake-AP chips, and indeed, its comparative numbers against that dual-socket Epyc 7601 system were obtained with SMT off on the AMD platform."
AMD schooled Intel.
When all you have is a hammer, every problem starts to look like a thumb.
This is another Skylake variant, plus an option of packing a quad system socket into a dual socket basically.
If you don't use the new SIMD instructions for INT8 and INT16 (neural network inference) a current quad socket wouldn't be too different, with a bit worse performance per watt (same silicon process as Kaby Lake on the older, same as Coffee Lake on the newer). Like, a current computer with four Xeon Gold 6000 series - I have no idea how big that is or how much that costs.