wtallis

33 ms·

631.
by wtallis 1y ago
> probably because the real hardware being used there is something like AVX and it can't be shared between all of the cores at the same time. That's not the right explanation; each physical core has its own vector ALUs for hand
632.
by wtallis 1y ago
That's eerily similar to what happened to Qualcomm's failed Snapdragon X Elite dev kit. That one eventually shipped in small quantities with a Type-C to HDMI dongle in the box to make up for the built-in HDMI port going missing. T
633.
by wtallis 1y ago
AMD has definitely been moving in that direction, and arguably doing a better job of it than Intel. But for now, AMD's desktop parts are still built with the same CPU core chiplets as their server parts, and none of the server parts ar
634.
by wtallis 1y ago
Die temperature readings aren't particularly helpful these days with desktop parts that will (depending on the power management settings) more or less keep increasing the clock speed until they reach ~90°C and just stay there. Upgradin
635.
by wtallis 1y ago
AMD's desktop CPUs are still running at a bit more than 1V; 1.3-1.4V is what you'll see at the high end of the clock speed range. But power draw can easily be in the 250–300W range if you turn on the "PBO" automatic over
636.
by wtallis 1y ago
Consider how some developers today talk about "toast" notifications: they understand on some level that calling them pop-ups is bad, but came to a consensus that using a different name was appropriate, rather than not actually mak
637.
by wtallis 1y ago
You're conflating two different things here: the performance metrics that marketing provides, and the product segments that marketing groups products into.
638.
by wtallis 1y ago
Intel and AMD have pretty much never made it straightforward to compare laptop part numbers against desktop part numbers and know what's faster or newer. That's not a new problem. If you compare the Ryzen AI Max+ 395 against the
639.
by wtallis 1y ago
The user (or bot) doesn't have to provide the hardware information, but does have to provide consent for that information to be sent to Valve. Bots have no strong reason to prefer "yes" or "no", but if you have to i
640.
by wtallis 1y ago
Samsung, SK Hynix, and Micron are the only three manufacturers doing leading-edge DRAM manufacturing, but there are other companies like Nanya that do older DRAM types and specialized DRAM parts for eg. embedded and industrial applications
641.
by wtallis 1y ago
There have been 48GB dual-rank DIMMs for about two years now, so 96GB using two slots and operating at high frequency has been an option. But even 64GB is still somewhat overkill for a gaming PC, putting you more into workstation territory.
642.
by wtallis 1y ago
The Steam Hardware Survey is an incredibly valuable resource, what with it being freely-available, constantly updated, and sourced from a population that makes its sampling biases generally easy to identify and understand. It and the Back
643.
by wtallis 1y ago
I think SK Hynix's NAND business may have already been bigger than Intel's NAND business when they made that acquisition. Certainly by then SK Hynix had recovered from being late to the 3D NAND transition, while Intel was on a wor
644.
by wtallis 1y ago
I don't think the touchbar was going to be worthwhile without haptic feedback. At the very least, it needed the force sensors used for the touchpad so that accidental touches could be properly rejected.
645.
by wtallis 1y ago
Based on https://learn.microsoft.com/en-us/windows/win32/fileio/spars... and subsequent links, it sounds like sparse files count against disk quotas as if they weren't sparse, but otherwise sparse f
646.
by wtallis 1y ago
It seems like just using a sparse file would make things simpler. I guess there would be potential inefficiency when you try to move the VM to another drive, but for just storing and using the VM in place it doesn't seem like this has
647.
by wtallis 1y ago
In my experience, nobody actually says "Tebibytes" out loud; it's just that silly. In writing, when the precision is necessary, the abbreviation "TiB" does see some actual use.
648.
by wtallis 1y ago
FYI, "TiB" means terabytes with a base of 1024, ie. the units you'd typically use for measuring memory rather than the units you'd typically see drive vendors using. The factor of 8 you divided by only applies to units b
649.
by wtallis 1y ago
Google also has the advantages that they now only need a read-only key-value store, and they know the frequency distribution for lookups. This is now the kind of problem many programmers would be happy to spend a weekend optimizing to get
650.
by wtallis 1y ago
> but usually compilers don't do that if you use the CPU baseline That's a problem that people are trying to solve by not using an ancient CPU baseline. Do you have a reasonable proposal for how else we should enable widespread
651.
by wtallis 1y ago
Its purpose is to serve as a repo for stuff that was never going to go on PyPI, whether or not PYX exists.
652.
by wtallis 1y ago
Aside from GDDR for GPUs, DRAM is still mostly specified with MT/s rather than GT/s, probably because marketing prefers bigger numbers. It'll probably fall off once 5-digit numbers become commonplace.
653.
by wtallis 1y ago
Am I missing something, or does SIMDe only help for cases where a program is using instruction intrinsics, and it doesn't do anything to address cases where the compiler decides to use SIMD as a result of auto-vectorization?
654.
by wtallis 1y ago
ffmpeg is a bad example, because it's the kind of project that has lots of infrastructure around incorporating hand-optimized routines with inline assembly or SIMD intrinsics, and runtime detection to dispatch to different optimized co
655.
by wtallis 1y ago
Microwave ovens top out around 1100-1250W output from a ~1500W input from the wall. Apparently there's a fair bit of energy lost in the power supply and magnetron that doesn't make it into the box where the food is.
656.
by wtallis 1y ago
> If not, there's no connection whatsoever between old hardware and unmaintained infrastructure. See, this is just you being unreasonable. Yes, we can all imagine why people might keep old hardware around. But your AlphaServer is at
657.
by wtallis 1y ago
It's hosted SaaS; you don't install it on your system.
658.
by wtallis 1y ago
One channel of DDR5-4800 actually competes pretty well against four channels of DDR3-1333 spread across two chiplets, which was the best Opteron configuration old enough to not have SSE4.1.
659.
by wtallis 1y ago
One server with today's tech can easily replace several servers that are 12+ years old. 4000 apps doesn't sound like a lot of work for one machine, unless you assume almost all of them are releasing new builds more than once a wee
660.
by wtallis 1y ago
OP did say "consumer Epyc", so presumably referring to the parts using the AM5 socket. From a quick check on Newegg, it looks like barebones servers for that platform start at under $1000, to which you need to add CPU, RAM, and st

More ›