4 ms·
I've been benchmarking these against Haswell and Broadwells. Despite being 300 MHz slower, we're getting between 5 and 45% faster benchmarks on linear algebra f
by zbjornson 10y ago
I've been benchmarking these against Haswell and Broadwells. Despite being 300 MHz slower, we're getting between 5 and 45% faster benchmarks on linear algebra functions that we run a lot, even without doing much work to tailor to AVX512 instructions yet.
The cache is also a whopping 56 MB.
- boulos 10y agoHi Zach! Are you using MKL at all under the hood (either through numpy, scipy, etc.) or is that just from AVX2 running much better clock-for-clock on Skylake? Disclosure: I work on Google Cloud.
- zbjornson 10y agoNo MKL, just a mix of AVX2 and a bit of AVX512 written with intrinsics in C++. Working on benchmarking Eigen at the moment, which is also AVX512-optimized.
- alfalfasprout 10y agoNot using Google Cloud but Intel just released MKL 2017 Update 2 today that automatically dispatches AVX512 code on supported Xeon CPUs. Doing a quick comparison between our Haswell servers and Skylake servers we do see a nice little speed bump of ~10-20% on matrix heavy code. The bump is negligible for most other things.
- eloff 10y agoStill limited to a measly 208gb of RAM? That was pretty good back in the day, but now my desktop workstation has 128gb. AWS offers up to 488gb on their latest R4 instances, and 2tb on their X1 instances released a year ago.
- boulos 10y agoFrom the November post [1]: > You'll be able to use the new processor with Compute Engine’s standard, highmem, highcpu and custom machine types. We also plan to continue to introduce bigger and better VM instance types that offer more vCPUs and RAM for compute- and memory-intensive workloads. [1] https://cloudplatform.googleblog.com/2016/11/power-up-your-Google-Compute-Engine-VMs-with-Intels-next-generation-Custom-Cloud-Xeon-Processor.html https://cloudplatform.googleblog.com/2016/11/power-up-your-G...
- crypt1d 10y agodamn. out of curiosity, what do u use that much ram for?
- jdhawk 10y agoDatabase
- warent 10y agoI honestly can't think of any viable purpose for that much RAM unless there's a huge amount of waste happening
- jsmthrowaway 10y agoHow hard are you thinking? I can think of numerous scenarios where an entire working set being in RAM is useful and occasionally essential, and it is quite common as an optimization in several industries. Finance comes to mind (kdb+ is designed to weaponize tons of RAM), as does high-scale Web operations. This is part of why solid state NVMe devices are compelling, too, to help bridge the gap, but it sounds like not for your purposes. Even on the less complicated end, some years ago throwing a shitload of page cache at MongoDB was the only way to maintain high write loads, and eventually one reached a point where one had to keep an entire shard in cache. That bottom end threshold is lower than you think. I don't know if that's changed.
- warent 10y agoI could be wrong but it really sounds like you're talking about production. If that's the case then yes, I can respect the RAM requirement and it's almost too obvious to state that it isn't purposeless. The key here is it's a personal development workstation on which he has these resources. If that's a requirement, I would argue there's something fundamentally wrong with his development processes requiring that much data locally without even a server.
- tiff_seattle 10y agoWe use multiple machines with 6 TB each at my job, all for use with databases.
- jdhawk 10y agoNo - I believe these will be up to 64core & 416GB
- iamgopal 10y agoI use 128 gb too but here we call it SSD