[X] My Assistant

Does anyone know if 128 ARM threads are useful? Because Twitter thought it was a good idea to recommend news about Ampere CPUs every 5 minutes

Does anyone know if 128 ARM threads are useful? Because Twitter thought it was a good idea to recommend news about Ampere CPUs every 5 minutes

128 threads could always be useful for <some workload.>
But in reality for most people there isn't much benefit. Most people's workloads aren't that parallel.

This post has been edited by dragontamer8740: Oct 27 2021, 06:27

I miss the pixiv of around six to ten years ago.

That site was a lot more usable and performant (even though at the time I thought it was sluggish, I had no idea what was to come).

And Pixiv Translation Plus was a great user script.

This post has been edited by dragontamer8740: Oct 22 2021, 06:17

Cannon created scanner/fax/printer combos that won't scan or fax if you run out of OEM ink. So its basically a subscription service.

[www.bleepingcomputer.com] https://www.bleepingcomputer.com/news/legal...run-out-of-ink/

Yeah I read about that.
Disgusting.

My newest canon product is from 1977 (an FD-mount 35mm SLR).
I have no intention of changing that, nor indeed of buying any currently manufactured consumer scanner/printer.

I'm hoping to some day get a large format SCSI scanner or something else no one wants but which is still of nice quality. As for printers… well, there's decades of them out there, and while the ink/toner is always stupid, I can find something from the last 20-25 years that can do "letter grade" (not dot matrix) printing and be happy for the most part. My motherboard still has a parallel port header on it that I can break out.

===Thought: random ramblings about SCSI===

Also, I always wondered why more varieties of peripherals didn't use SCSI. If scanners could use it, then one would think other kinds of devices could… video capture cards/framegrabbers come to mind. Since you can daisy chain between 8 and 16 devices to a bus, it seems pretty handy.

And if manufacturers didn't cheap out on ground pins like Apple's DB25 SCSI connector, they were pretty reliable, too.
I think if the standard had either required all devices to be able to auto-terminate, or required that termination not be integrated into devices (e.g. requiring a discrete terminator piece), the standard would become nearly idiot-proof (I'm ignoring LVD and HVD "differential SCSI," and just talking about the 5 volt variety).

Yeah, ATA drives and controllers would still be cheaper to implement, but as an external high speed bus I'd say it's a hell of a lot better than a parallel port or RS-232 for 90s consumer stuff. I'm imagining a 90's non-apple consumer laptop that would have a SCSI connector to attach stuff to, basically outmoding the serial mouse and "backpack" parallel port CD drives and acting as a plug-n-play predecessor to USB (and ultra320 works at much higher speeds than USB 1.x, 2.x, or even "Firewire 800". Even SCSI-1 from the 80's could outspeed USB 1.1).

I still think RS232 is brilliant in its simplicity and for low speed stuff, though. Only thing I'd have changed is making it 5V TTL instead of ±12V, but 24Vp-p RS-232 still allows really great cable lengths.

This post has been edited by dragontamer8740: Oct 22 2021, 08:34

Some stuff, especially cheap converter dongles don't do real RS-232 voltage levels- which works fine till you find something that actually demands proper levels and handshaking be present. This has caused me a few headaches. There's also the modern widgets which use an explicitly 5V or 3V3 TTL bus with no handshaking, but otherwise RS-232 style signalling. Mercifully, these don't typically include a DE-9 and just use header pins, TRS, or some such.

Fan on my raid card died yesterday. Ordered a less generic 40x10mm 3 pin fan. I'll probally chop the cable down and cap off the 3rd wire, as I have just a 2-pin header setup to power it.

Sweet hentai-themed stickers on Telegram

Power supply swapped out in big computer and new fan installed on raid card. Had to make a custom cavke with a molex male to replace a custom one on the old supply- two cable mounted fan headers, and a long molex female to reach down to the serial card.
After having a flase alarm on a drive failure and being unabke to silence the loud annunciator, finally got the MegaRAID management software working, so now it's trivial to configure the card and read logs.

The smaller Xeon box is working well enough now, finally got the mobo use my graphics card. Turns out there's a bios option for boot display adapter. Somebody on IRC told me how to get both display adapters working at once if I'm so inclined, and how to setup the bios for best performance on the CPU. (Not that interested tbh, as I'm using daisy chained adapters to vuew the VGA, and something in the chain doesn't suport the native mode of my monitor, so the resulting video is quite bad; I'll just conside the VGA a debug tool.)
He also had info for a linux program to access the management geatures of the mobo, and a little tweak to make linux run better on haswell xeon. Going to do a clean install eventually instead of using the linux drive/install that's been used on 4 different mobos now- lots of things aren't quite right, and it's slower than it ought to be.
Got USB 3 retrofitted, and changed out the front panel's power LED to a soft green instead of it's prior blue. It was the only SMD one on the board, but I was abke to dead-bug a 3mm LED in it's place.

Said IRC guy has a dual 32-core xeon rig, with supermicro X11 series motherboard, running hackintosh.

If not in mac os you can usually set custom resolutions somewhat painlessly.
On Mac OS you usually have to construct an EDID, as I've mentioned in the past.
X11/Unix example for 1280x800 @ 60hz:

CODE

$ cvt 1280 800 60
Modeline "1280x800_60.00"   83.50  1280 1352 1480 1680  800 803 809 831 -hsync +vsync

copy everything after but not including the word 'Modeline' and then paste it as below:

CODE

xrandr --newmode "1280x800_60.00"   83.50  1280 1352 1480 1680  800 803 809 831 -hsync +vsync

Find out what your outputs are called:

CODE

$ xrandr
Screen 0: minimum 8 x 8, current 1280 x 800, maximum 32767 x 32767
LVDS1 connected primary 1280x800+0+0 (normal left inverted right x axis y axis) 260mm x 160mm
---snip---
DP1 disconnected (normal left inverted right x axis y axis)
HDMI1 disconnected (normal left inverted right x axis y axis)
VGA1 disconnected (normal left inverted right x axis y axis)

Then run the following to associate your new resolution/mode with the output you have (I'm using VGA1; also, your mode name should be whatever was at the begining of the string you pasted into --newmode. You can change the name before running the newmode command and it'll work, too, as long as the name is consistent):

CODE

xrandr --addmode VGA1 1280x800_60.00

Then switch to the new resolution:

CODE

xrandr --output VGA1 --mode 1280x800_60.00

'cvt' stands for 'Coordinated Video Timing.' There's a similar tool called 'GTF' ('Generalized Timing Formula') that will give slightly different results and may be worth a try if your display doesn't sync to the mode given by cvt. Both are VESA standards; CVT is the newer one.

For boot time, you can specify resolution in grub configuration using 'GRUB_GFXMODE' and 'GRUB_GFXPAYLOAD_LINUX=keep' to preserve that resolution for the early boot process. On (U)EFI systems, disabling the CSM in your bios settings can help with using high resolutions via 'efifb.'

This post has been edited by dragontamer8740: Oct 25 2021, 20:16

The chained adapters manage to pass 1920x1080 60Hz instead of 1920x1200 75Hz. There's also some overscan and a faint moving diagonal artifact.
Ultimately, I'd just use the video card proper for a second display, unless I needed VGA instead of digital video for something.

Thanks for the tip in any case.

Tried to get back into MVVM programming (I don't know on what project I tried it before, but can't find it somehow).
Damn hell of spending an absurd amount of way on the absurd GUI with the absurd bugs and the absurd logic...well that should not be the work of a single person, even just for the view side. (IMG:[invalid] style_emoticons/default/dry.gif)

The future of technology are smart cores (XeSS, Google Tensor and Nvidia RTX) or brute force (like Intel, AMD, Qualcomm and partially Apple)?

The future of technology are smart cores (XeSS, Google Tensor and Nvidia RTX) or brute force (like Intel, AMD, Qualcomm and partially Apple)?

Maybe, they do certain workloads extremely well but it taking the silicon space from traditional processing.


Great for the AI lead dystopia.

the future is banging rocks together after the revolution comes

Anime tiddies carved into stone tablets.

Some stuff, especially cheap converter dongles don't do real RS-232 voltage levels- which works fine till you find something that actually demands proper levels and handshaking be present. This has caused me a few headaches.

My IBM terminal requires hardware handshaking signals, but it seems lenient on levels. I think it might require it to at least swing negative (no DC offset/bias), but it seems content to accept whatever my adapter is pushing out (I think it's probably 10Vpp using a charge pump or some such).
Apparently, the VT-100 didn't even strictly require hardware handshaking. Just Rx/Tx. Probably runs a lot more smoothly with it though. IDK, I have never actually had the pleasure of seeing a DEC terminal of any sort.

Fan on my raid card died yesterday. Ordered a less generic 40x10mm 3 pin fan. I'll probally chop the cable down and cap off the 3rd wire, as I have just a 2-pin header setup to power it.

You saw the pic of my old Galaxy GT630 card a while back, right? I did a similar deal there but I had a 2-pin fan on a 4-pin connector.

It was probably really shit for airflow though, since it had those fins sticking off around it presumably for air to blow out the sides. But since it was a GT630 I think that might have been overkill.

The one really nice thing about that card was that it only took a single slot.

This post has been edited by dragontamer8740: Oct 26 2021, 18:48

Maybe, they do certain workloads extremely well but it taking the silicon space from traditional processing.
Great for the AI lead dystopia.

Then it's time to go back to the old reliable: Dual Processor, or rather, Dual SoC

Wondering if it'd be worthwhile to upgrade small computer's PSU. It has a 465W bronze, and according to a [outervision.com] power supply calculator, my suppy is rather marginal- as in it's reccomendation is about 3W higher. The supermicro instead of consumer grade supply makes a bit better about it, but meh. Also mulled over an E3-1281 v3, but it'd just be a very marginal clock increase. The only E3 Broadwell Xeons are low power chips, and an APU one that seems worse on paper in every way.


The little thing with the blue plug behind my glasses is the remote for the KVM.

The red "i" light is blinking to indicate fan failure, because my low speed 92x15mm Scythe CPU fan, and 80x25mm Noctua front fan are dipping below 300RPM at idle. 300RPM is hard coded into the bios to indicate fan failure. Noctua sells little in-line fan controllers that have a 300RPM minimum mode just for this, but they're $21 a pop. Could also look into higher RPM Nidec or Sanace fans like it'd have OEM. That'd cost even more though, the exact OEM front fan is $44 from the only source I can find. That said, I can likely find fans for about the same price...

This post has been edited by Wayward_Vagabond: Oct 28 2021, 06:57

Does anyone know if 128 ARM threads are useful? Because Twitter thought it was a good idea to recommend news about Ampere CPUs every 5 minutes

For very specific work loads. Its why people are building cluster Pi4 builds.

The Ampere cpus will be good for certain workloads. They are talking about cloud applications and white box servers.

The red "i" light is blinking to indicate fan failure, because my low speed 92x15mm Scythe CPU fan, and 80x25mm Noctua front fan are dipping below 300RPM at idle. 300RPM is hard coded into the bios to indicate fan failure. Noctua sells little in-line fan controllers that have a 300RPM minimum mode just for this, but they're $21 a pop. Could also look into higher RPM Nidec or Sanace fans like it'd have OEM. That'd cost even more though, the exact OEM front fan is $44 from the only source I can find. That said, I can likely find fans for about the same price...

You could use a [en.wikipedia.org] frequency multiplier to report doubled tach speeds.

I'd try a PLL-based method, but whatever works.

This post has been edited by dragontamer8740: Oct 28 2021, 04:38