|
BangersInMyKnickers posted:why are you replacing server gear with consumer grade garbage? the best part about server gear is that it's never obsolete.
|
# ? Nov 26, 2019 16:44 |
|
|
# ? May 8, 2024 16:40 |
|
akadajet posted:the best part about server gear is that it's never obsolete. not as long as it's under warranty
|
# ? Nov 26, 2019 16:45 |
|
Farmer Crack-rear end posted:lol i looked up the hp k-class online and a few hops later found out that there was a product named "HP Superdome" they still make it (kinda) its called the superdome flex 💪 now
|
# ? Nov 26, 2019 18:20 |
|
i saw someone propose hooking a fully kitted out flex to that cerebras ml megachip
|
# ? Nov 26, 2019 18:23 |
|
anybody have an example of what a two socket epyc rome server is not good at?
|
# ? Nov 26, 2019 18:26 |
|
PCjr sidecar posted:anybody have an example of what a two socket epyc rome server is not good at? intel is faster for ai if you use their "deep learning boost" instructions https://arstechnica.com/gadgets/2019/11/hands-on-with-amds-32-core-64-thread-threadripper-3970x/ but i think that same work could be done cheaper faster on a gpu
|
# ? Nov 26, 2019 18:51 |
|
my ipad has a NEURAL ENGINE
|
# ? Nov 26, 2019 21:02 |
|
infernal machines posted:it needed to sell in louisiana i'm not a louisian, or former louisian as hp is, but i don't believe the superdome has a particularly good reputation in louisiana? i guess hp probably has a quip about it
|
# ? Nov 26, 2019 21:38 |
|
DELETE CASCADE posted:my ipad has a NEURAL ENGINE my Linux rig has an emotion engine just kidding I don’t run Linux lol
|
# ? Nov 26, 2019 23:47 |
|
xD
|
# ? Nov 27, 2019 08:17 |
|
BangersInMyKnickers posted:why are you replacing server gear with consumer grade garbage? i love server grade gear https://support.hpe.com/hpsc/doc/public/display?docId=emr_na-a00092491en_us quote:SSDs with an HPE firmware version prior to HPD8 that results in SSD failure at 32,768 hours of operation (i.e., 3 years, 270 days 8 hours). After the SSD failure occurs, neither the SSD nor the data can be recovered. In addition, SSDs which were put into service at the same time will likely fail nearly simultaneously.
|
# ? Nov 27, 2019 09:37 |
|
wish this thread would
|
# ? Nov 27, 2019 11:46 |
|
I dunno why people are janitoring most ryzen stuff since boost clocks hit basically what the chip can handle under any normal cooling solution
|
# ? Nov 27, 2019 19:32 |
|
my stepdads beer posted:i love server grade gear wow, i can't believe stymie was wrong about enterprise ssds, too
|
# ? Nov 27, 2019 21:57 |
|
Fabricated posted:I dunno why people are janitoring most ryzen stuff since boost clocks hit basically what the chip can handle under any normal cooling solution epeen. I have my 9700k at 5.0 stable because... well gently caress, I can, and that speed is near-mythical for me since I remember a time when that was functionally impossible.
|
# ? Nov 28, 2019 01:02 |
|
Farmer Crack-rear end posted:intel also had Slot 2 for the Xeon processors, i've got a couple of new-in-box P2 xeons squirreled away somewhere in my mom's house. those suckers were big why are you hoarding garbage in someone else’s house
|
# ? Nov 28, 2019 04:31 |
|
and why not in my house. i want that poo poo
|
# ? Nov 28, 2019 09:42 |
|
infernal machines posted:iirc their speculative execution implementation wasn't vulnerable to spectre/meltdown to the same extent intel's was so they didn't have to take the performance hit to fix it, while intel did Not just threadrippers, but regulary Ryzens too...
|
# ? Nov 28, 2019 10:32 |
|
Fabricated posted:I dunno why people are janitoring most ryzen stuff since boost clocks hit basically what the chip can handle under any normal cooling solution people want to feel like they are getting more than their moneys worth by squeezing some marginal extra performance amd are happy to oblige
|
# ? Nov 28, 2019 10:51 |
|
Fabricated posted:I dunno why people are janitoring most ryzen stuff since boost clocks hit basically what the chip can handle under any normal cooling solution the first gen took some janitoring because r7 1700 boosting goes 1 or 2 core load = 3.7GHz and any more than that nothing goes over 3.2GHz which is kinda poo poo though all it took was setting clocks to 3.8GHz and bumping the voltage up slightly and then never touching it again
|
# ? Nov 28, 2019 17:10 |
|
intel is bad
|
# ? Nov 28, 2019 22:46 |
|
there can only be one good
|
# ? Nov 28, 2019 23:16 |
|
crazy to think processors are close to or already small enough for quantum effects to interfere
|
# ? Nov 29, 2019 01:37 |
|
Technically, everything is influenced by quantum effects.
|
# ? Nov 29, 2019 01:49 |
|
what I mean is that older larger processors didn’t really have to consider them in the design process but I believe now you do, with the 10nm and smaller processes I think
|
# ? Nov 29, 2019 01:51 |
|
echinopsis posted:what I mean is that older larger processors didn’t really have to consider them in the design process source?
|
# ? Nov 29, 2019 03:06 |
|
quantum tunneling is a thing that needs to be designed around afaik
|
# ? Nov 29, 2019 04:46 |
|
quantum tunneling is also a factor at much larger sizes fwiw. it’s a matter of probability whether an electron hops the barrier between leads, and the probabilities drastically increase as distances shrink, but they can technically happen at any distance. as a bonus the smaller scale also makes your hardware much more sensitive to individual electrons getting into places they shouldn’t be im not a hardware dev but I had a lot of physics classes a decade ago
|
# ? Nov 29, 2019 05:29 |
|
nand flash is literally applied quantum tunneling
|
# ? Nov 29, 2019 07:09 |
|
The Management posted:nand flash is literally applied quantum tunneling explain i'm a baby and physics scares me
|
# ? Nov 29, 2019 08:24 |
|
SRQ posted:explain i'm a baby and physics scares me basically, you have a totally isolated bit of metal that's electrically insulated from from everything. you use quantum tunneling to either stick some electrons in there or take them back out, and once you do that they're stuck there because it's totally electrically insulated. when you want to read the value of that bit, you measure whether there are extra electrons stuck on that floating bit of metal without actually giving them any chance to escape.
|
# ? Nov 29, 2019 08:36 |
|
that rules
|
# ? Nov 29, 2019 08:59 |
|
Semiconductors themselves rely on quantum physics, it's just specific effects which become more important at smaller sizes.
|
# ? Nov 29, 2019 09:09 |
|
my stepdads beer posted:that rules
|
# ? Nov 29, 2019 09:59 |
|
pointsofdata posted:Semiconductors themselves rely on quantum physics, it's just specific effects which become more important at smaller sizes. but enough about your sex life! haha!
|
# ? Nov 29, 2019 10:07 |
|
Jabor posted:basically, you have a totally isolated bit of metal that's electrically insulated from from everything. you use quantum tunneling to either stick some electrons in there or take them back out, and once you do that they're stuck there because it's totally electrically insulated. but how do you use quantum tunneling to do that
|
# ? Nov 29, 2019 11:45 |
|
pointsofdata posted:Semiconductors themselves rely on quantum physics, it's just specific effects which become more important at smaller sizes. This is kind of true, but is often overstated, IMO. To understand the physical theory for why semiconductors are semiconductors, yes, you need to be quantum mechanical, but semiconductor electronics isn't really that quantum mechanical. The main physics of transistors is just normal physics. SRQ posted:but how do you use quantum tunneling to do that I don't know if the following explanation will be helpful, but: The isolated piece of metal Jabor is talking about is the bottom plate of a capacitor. You apply a large voltage to the top plate of this capacitor with an electrical insulator separating the top plate and bottom plate. This large voltage effectively thins this insulator which usually prevents electricity from traveling from the top plate to the and bottom plate of the capacitor, and then electrons can now travel through the temporarily 'effectively thinned' insulator layer from the top plate to the bottom plate. The process of the electrons traveling through the insulator is forbidden/disallowed by laws of normal physics, for the same reason why human beings cannot walk through walls, but in quantum mechanics it can happen with a very high probability if the insulator/wall is sufficiently 'effectively thin'. The bottom plate of this capacitor is connected to the input of a transistor, which is essentially a charge sensing device which detects if a sufficient number of electrons have made it to the bottom plate of this capacitor. When a memory read is performed, the memory circuit then reads out from the transistor output whether a sufficient number of electrons have been introduced to the bottom plate of the capacitor. silence_kit fucked around with this message at 12:39 on Nov 29, 2019 |
# ? Nov 29, 2019 12:21 |
|
silence_kit posted:This is kind of true, but is often overstated, IMO. To understand the physical theory for why semiconductors are semiconductors, yes, you need to be quantum mechanical, but semiconductor electronics isn't really that quantum mechanical. The main physics of transistors is just normal physics. It's closer under the surface than in most physics though, "Why does semiconductor conductivity increase with temperature?" quickly gets you to quantum physics.
|
# ? Nov 29, 2019 14:02 |
|
how i want a drink, alcoholic of course, after the heavy lectures involving quantum mechanics that's a pi mneumonic
|
# ? Nov 29, 2019 15:57 |
|
|
# ? May 8, 2024 16:40 |
|
too bad everyone ran off the quantum math c++ library guy. I am sure he would have had some great informative posts here
|
# ? Nov 29, 2019 16:17 |