|
when has pcie saturation ever, ever, been the primary or even secondary system bottleneck?
|
#
?
Feb 7, 2016 03:20
|
|
|
|
| # ? May 17, 2024 05:05 |
|
|
thats our nbsd 
|
|
#
?
Feb 7, 2016 03:54
|
|
|
there are plenty of cases and studies that show most computer issues are memory problem so ecc is indeed good, but i never heard of pcie ever being a problem
|
|
#
?
Feb 7, 2016 04:25
|
|
|
ive used a lot of compys and my middling tier dollar per pwn from the parts picker thread have been the best by a large margin since forever. hard to believe how much ownage the shsc boys can serve up in that drat OP
|
|
#
?
Feb 7, 2016 08:32
|
|
|
ive used em all baby. free netbooks, chromebooks, alienwars, ASUS ROG superlappys, and even a murdered out 17" MPB or whatever the most expensive one was at the time, unibody aluminum c2d. and at the end of the day lads, when i need a compy that can do anything - that has no limits. the compy to *home depot guy voice* Get Things Done. it's all about the beefy gamer rig. Does it all folks.
|
|
#
?
Feb 7, 2016 08:34
|
|
|
one time my gamer rig experienced a catastrophic failure when the power supply died. went to frys, 1 hour later? back in business. easy. GG
|
|
#
?
Feb 7, 2016 08:35
|
|
|
Smythe is a full sick hacking genius, so don't mess
|
|
#
?
Feb 7, 2016 16:32
|
|
|
Fuzzy Mammal posted:when has pcie saturation ever, ever, been the primary or even secondary system bottleneck? the moment you want your nic and your nvme to work at the same time
|
|
#
?
Feb 7, 2016 18:06
|
|
|
Celexi posted:there are plenty of cases and studies that show most computer issues are memory problem so ecc is indeed good, but i never heard of pcie ever being a problem come for the ecc, stay for the i/o bandwidth!
|
|
#
?
Feb 7, 2016 18:07
|
|
|
I play video games on computer
|
|
#
?
Feb 7, 2016 20:17
|
|
|
I made computer myself!
|
|
#
?
Feb 7, 2016 20:17
|
|
|
BONGHITZ posted:I play video games on computer BONGHITZ posted:I made computer myself! 2 wrongs don't make a right
|
|
#
?
Feb 7, 2016 20:24
|
|
|
Notorious b.s.d. posted:come for the ecc, stay for the i/o bandwidth! i'll say this, as a ramdisk using memory whore, i didn't notice that huge of a hit to write speeds with ecc, and i think that's about the most memory intensive application i can think of outside of some really esoteric & fucky computation
|
|
#
?
Feb 7, 2016 21:36
|
|
|
lol ramdisks compute like it's 2005
|
|
#
?
Feb 7, 2016 22:12
|
|
I CANNOT EJACULATE WITHOUT SEEING NATIVE AMERICANS BRUTALISED!
|
blowfish posted:lol ramdisks
|
|
#
?
Feb 7, 2016 22:15
|
|
|
between win10 and samsung magician ramdisks are pretty much done for you
|
|
#
?
Feb 8, 2016 01:29
|
|
|
|
|
#
?
Feb 8, 2016 09:13
|
|
|
surebet posted:i'll say this, as a ramdisk using memory whore, i didn't notice that huge of a hit to write speeds with ecc, and i think that's about the most memory intensive application i can think of outside of some really esoteric & fucky computation ecc doesn't hurt throughput. it hypothetically hurts latency. if it normally takes, say, 10 clock cycles for your memory to start streaming your request back to you, ecc makes it 11 cycles. the bandwidth is the same, there's just that extra clock cycle spent waiting. i just looked up what the exact loss would be, and it turns out that it's negative, because the registered memory is so much more tightly timed than unbuffered, it overcomes the ecc gap DDR4 2400 R (registered, always ECC) runs at a standard timing of 16 cycles DDR4 2400 U (unbuffered, rarely ECC) runs at a standard timing of 18 cycles so if you use registered ECC memory it is actually a few percent faster than standard memory
|
|
#
?
Feb 21, 2016 05:52
|
|
|
that said cheapo 1-socket "workstation" systems are not gonna support registered memory so your unbuffered ecc will be 19 cycles instead of 18 cycles. your memory latency will be 5.55% higher! your pipeline stalls on memory access will be 36 cycles + 60 ns, instead of 36 cycles + 57 ns! gently caress
|
|
#
?
Feb 21, 2016 05:58
|
|
|
legit question notorious bsd straight from intel ark, one of the cheapest pentiums you can buy (like $86 or something)  no ecc support?
|
|
#
?
Feb 21, 2016 06:47
|
|
|
it seems some newer pentiums and i3 support ecc, which is nice as only amd had ecc on regular cpus
|
|
#
?
Feb 21, 2016 06:50
|
|
|
Wild EEPROM posted:legit question notorious bsd those are sold for use in very low-end servers. like, 1u rackmount fileservers for branch offices. poo poo where you would literally use an atom if they came with ecc. again, intel's goal is price segmentation. they want to extract the maximum price from each and every individual user. there's a group of extremely price-sensitive users that requires ecc but gives zero fucks about performance. intel has a product and price point specificaly targeting this group
|
|
#
?
Feb 21, 2016 07:11
|
|
|
lmao 
|
|
#
?
Feb 21, 2016 07:12
|
|
|
pram posted:lmao
|
|
#
?
Feb 21, 2016 15:52
|
|
|
|
| # ? May 17, 2024 05:05 |
|
|
Notorious b.s.d. posted:an atom if they came with ecc. those exist
|
|
#
?
Feb 21, 2016 16:50
|
|