« first day (3144 days earlier)      last day (1890 days later) » 

12:58 AM
Catching up on a few days' of tech news...
(still on this cheap laptop, now that I have the benefit of zram)
 
 
1 hour later…
2:00 AM
> In addition to its flexible power management, the PBlaze5 supports several of the more advanced NVMe features that are often left out on entry-level enterprise drives. The drive supports 128 NVMe queues, so all but the largest servers will be able to assign one queue to each CPU core, allowing IO to be performed without core to core locking or synchronization. Many older enterprise NVMe SSDs we have tested are limited to 32 queues, which is less than optimal for our 36-core testbed.
I wonder how many queues consumer NVMe SSDs support...
Intel's old SSD 750 Series does 32.
Toshiba OCZ RD400: 7 queues.
Intel SSD 600p: 8 queues.
 
 
1 hour later…
3:26 AM
Feb 22 at 23:20, by bwDraco
GeForce GTX 1660 Ti: 1536 CUDA cores in 24 of 24 SMs (fully enabled) on Turing TU116 @ 1500-1770 MHz; 6 GB GDDR6 @ 12 Gbps on 192-bit bus; 5.5 GFLOPS FP32; TSMC CLN12FFN process, 120W TDP. No RTX or tensor cores.
GeForce GTX 1660: 1408 CUDA cores in 22 of 24 SMs on Turing TU116 @ 1530-1785 MHz; 6 GB GDDR5 @ 8 Gbps on 192-bit bus; 5.0 TFLOPS FP32; TSMC CLN12FFN process, 120W TDP.
Not too sure why the original message had GFLOPS and not TFLOPS...
So... we lose only a slight amount of shader performance but memory speeds see a serious cut.
The higher clock frequency mostly makes up for the two fewer SMs, but the memory bandwidth will hurt performance more noticeably, especially at high resolutions. This is best seen as a competitor to the RX 590 for 1080p gaming.
I would not be surprised if the majority of the cost difference from the Ti card is due to the memory. Slower 8 Gbps GDDR5 memory is probably much cheaper than GDDR6, especially when 10+ Gbps GDDR5X was broadly used at the high end of the Pascal lineup.
 
 
2 hours later…
Bob
5:46 AM
probably G+
 
Ooh. That is going to be 'fun'
 
6:39 AM
Huh...
I wonder if I have a G+ avatar anywhere.
Most G+ logons would be just Google now... right?!
 
7:25 AM
What are you trying to tell me Unity?
 
 
6 hours later…
1:36 PM
@ThatBrazilianGuy i have that! would love to race against some of you
 
2:00 PM
GRiD2.get()
 
 
4 hours later…
6:25 PM
Patch panel or keystone? flips coin
 
hm
where?
patchpanel makes sense for centralised wrangling
keystone makes sense for rooms
 
This hallway thingymabob
I've only got 5 to terminate, so it's much of a muchness
Haven't yet decided whether to go the full hog and buy a mini comms cupboard (this CableMoneky one is nice but also £87 quid), or just mount a panel on a couple of rails / mounts screwed to the wall (± boxed in with some wood)
On the one hand the keystones have the advantage that I can finally terminate the cables lying in a tangled mess in my hallway and decide what to do later, and the disadvantage that I probably would just leave them in that state for another $ages ;-P
 
7:09 PM
Keystones it is!
 
 
2 hours later…
9:10 PM
@bertieb "8U" Looks like it might fit, what... 4? And that's if you stand them on their ends.
 
@MichaelFrank Small aye but I don't really intend on putting much in there :)
 
 
1 hour later…
10:30 PM
a friend sent me this:
> The value of the number pi has been calculated to a new world record length of 31 trillion digits, far past the previous record of 22 trillion.

The calculation required 170TB of data (for comparison, 200,000 music tracks take up 1TB) and took 25 virtual machines 121 days to complete.

It would take 332,064 years to say the 31.4 trillion digit number.
why would they use VMs for this?
surely physical machines would be faster otherwise you'd run into CPU contention issues?
 
11:10 PM
Cloud servers.
 
11:20 PM
roar
Five hours of baseball photography.
 
even then you'd hit contention issues though, right?
 

« first day (3144 days earlier)      last day (1890 days later) »