|
Beve Stuscemi posted:
the xbox 360 is still the reigning king there
|
# ? Apr 20, 2024 17:46 |
|
|
# ? May 14, 2024 08:57 |
|
true. ironically the ps3 gives it a run for its money
|
# ? Apr 20, 2024 18:11 |
|
Pendragon posted:along those lines: elaborate cooling instructions involving shaving the top off your CPU and precisely spreading thermal paste with a razor back in the day when it theoretically might have helped but was more likely to destroy your CPU entirely vs. now where you just blob the stuff in a pattern, clamp it the hell down and you're done
|
# ? Apr 20, 2024 20:41 |
|
hm so it looks like "kegabyte" wasn't actually a thing in the early 90s, i assume i'd extrapolated this from mb being "megabyte" when i was a wee lad
|
# ? Apr 20, 2024 20:43 |
|
kegalbyte?
|
# ? Apr 20, 2024 20:46 |
|
Kitfox88 posted:kegalbyte? No vagina dentata here tyvm
|
# ? Apr 20, 2024 21:24 |
|
Progressive JPEG posted:hm so it looks like "kegabyte" wasn't actually a thing in the early 90s, i assume i'd extrapolated this from mb being "megabyte" when i was a wee lad maybe you just conflated kilobyte and megabyte? most files weren't even megabytes back then, remember a floppy was usually at most 2 MB i still don't use mebibyte and gibibyte etc. because they sound stupid lol
|
# ? Apr 20, 2024 22:04 |
|
people actually taking a portable computer off the desk, sometimes
|
# ? Apr 20, 2024 22:09 |
|
3D Megadoodoo posted:people actually taking a portable computer off the desk, sometimes i do this all the time
|
# ? Apr 20, 2024 22:13 |
|
insane
|
# ? Apr 20, 2024 22:15 |
|
Beeftweeter posted:i still don't use mebibyte and gibibyte etc. because they sound stupid lol i never will a megabyte will always be 1024 kilobytes. fight me, IEC
|
# ? Apr 20, 2024 22:20 |
|
mebibyte, mebinot
|
# ? Apr 20, 2024 22:27 |
|
nobody has ever used or will ever use that kibi/mebi/gibibyte poo poo except for wikipedia weirdos
|
# ? Apr 20, 2024 22:32 |
|
i think one of the standards bodies made a (not terribly big) push for using it around maybe 1999-2000 but i can't remember which one (it is 4/20 after all). but i do clearly remember laughing at it and saying "yeah gently caress that" e: it might have been slightly later, like maybe 2001-2 at the latest, but it predated wikipedia being popular regardless
|
# ? Apr 20, 2024 22:58 |
|
Powerful Two-Hander posted:elaborate cooling instructions involving shaving the top off your CPU and precisely spreading thermal paste with a razor back in the day when it theoretically might have helped but was more likely to destroy your CPU entirely oh god I spent far too long lapping my first heatsink when it was smooth to begin with geez I was an idiot
|
# ? Apr 20, 2024 23:19 |
|
Powerful Two-Hander posted:elaborate cooling instructions involving shaving the top off your CPU and precisely spreading thermal paste with a razor back in the day when it theoretically might have helped but was more likely to destroy your CPU entirely
|
# ? Apr 20, 2024 23:26 |
|
for a laugh look up the TDP's of the old hot and slow netburst pentium 4 chips and compare 'em to the TDP's of modern ryzen and core CPU's
|
# ? Apr 21, 2024 02:11 |
|
i don't know much about how cpus work in extreme detail tbh, but from my understanding wasn't netburst designed to be, i guess to put it simply, generally better at higher frequencies than lower ones? i know that meant they had crazy high TDPs, but i wonder what would happen if someone made a modern chip on a modern process, based off of the netburst concept (i'd guess just using the same design would probably be bad) and clocked it real high. like, maybe the idea was valid, but we just couldn't manufacture it at the time? or maybe it just sucked entirely, idk lol. i remember some revisions being pretty good, like northwood
|
# ? Apr 21, 2024 02:33 |
|
Beeftweeter posted:i don't know much about how cpus work in extreme detail tbh, but from my understanding wasn't netburst designed to be, i guess to put it simply, generally better at higher frequencies than lower ones? theoretically for chip architecture reasons a longer pipeline makes it easier to reach higher clock speeds because each stage is doing less work, or something. but netburst had a long rear end pipeline, like prescott had a ridiculous 31 stage pipeline, so you blast away a ton of work in the pipeline on a branch misprediction. so to make up for that design flaw they were like, let's make the branch predictor really good and jack the clock rate really high so it doesn't matter quote:i know that meant they had crazy high TDPs, but i wonder what would happen if someone made a modern chip on a modern process, based off of the netburst concept (i'd guess just using the same design would probably be bad) and clocked it real high. like, maybe the idea was valid, but we just couldn't manufacture it at the time? it just sucked lol there's a reason they threw it in the trash
|
# ? Apr 21, 2024 04:06 |
|
for a while i had a dual netburst xeon server with four 15k SAS drives and to this day im surprised i didnt die in that room of heatstroke
|
# ? Apr 21, 2024 04:08 |
|
the big idea of netburst was "moar clocks". like the whole point was to make a cpu that can clock as high as possible - they were hitting 4GHz on a 90nm process, and predicting 10GHz by 2010. that's impressive as hell, but it's not the metric that actually matters for real-world performance - instructions-per-clock is equally important (which netburst sucked at in practice), and in many modern applications instructions-per-watt is the real key. to reach those clocks you need a long pipeline that only does a tiny amount of work at each step - you need the signal to propagate from the start of each step to the end before the next clock tick - so either you eat poo poo on every single branch or you dedicate immense amounts of your chip area to branch prediction and caching to try and avoid that. netburst did dedicate immense amounts of chip area to that, and still ate poo poo every time your code had branches in it (even successfully predicted branches!) that went over what it could fit in the branch trace cache (so, literally anything bigger than a benchmark, essentially). meanwhile amd was kicking its rear end with athlon by using that area for moar instruction and data cache.
|
# ? Apr 21, 2024 04:14 |
|
Beeftweeter posted:i think one of the standards bodies made a (not terribly big) push for using it around maybe 1999-2000 but i can't remember which one (it is 4/20 after all). but i do clearly remember laughing at it and saying "yeah gently caress that" the main thing i remember them being pushed was as an explanation for why hard drive sizes were consistently smaller than advertised. "oh no see when we say 100 gigabytes we mean 100,000,000,000 bytes " poo poo. disk sizes are a cluster gently caress.
|
# ? Apr 21, 2024 06:11 |
|
1.44 MB floppies are neither 1.44 * 1,000,000 bytes nor 1.44 * 1,048,576 bytes they're 1.44 * 1024 * 1000 bytes
|
# ? Apr 21, 2024 08:19 |
|
shackleford posted:1.44 MB floppies are neither 1.44 * 1,000,000 bytes nor 1.44 * 1,048,576 bytes ... But why?
|
# ? Apr 21, 2024 08:21 |
|
Volmarias posted:... But why? i guess because it's twice the size of a 720 KB floppy
|
# ? Apr 21, 2024 08:28 |
|
the first retail 4ghz intel chip wasnt netburst even though it was trivial to overclock well past that it was the haswell era 4790k like 5 years later
|
# ? Apr 21, 2024 08:40 |
|
remember how the first pentium 4s used Rambus RAM? just lol
|
# ? Apr 21, 2024 14:50 |
|
shackleford posted:theoretically for chip architecture reasons a longer pipeline makes it easier to reach higher clock speeds because each stage is doing less work, or something. but netburst had a long rear end pipeline, like prescott had a ridiculous 31 stage pipeline, so you blast away a ton of work in the pipeline on a branch misprediction. so to make up for that design flaw they were like, let's make the branch predictor really good and jack the clock rate really high so it doesn't matter Jabor posted:the big idea of netburst was "moar clocks". like the whole point was to make a cpu that can clock as high as possible - they were hitting 4GHz on a 90nm process, and predicting 10GHz by 2010. that's impressive as hell, but it's not the metric that actually matters for real-world performance - instructions-per-clock is equally important (which netburst sucked at in practice), and in many modern applications instructions-per-watt is the real key. huh i see. so even if the idea were valid (which it seems it wasn't, other than for squeezing out higher clock speeds on big-for-today processes) it wouldn't be any good at a modern workload anyway kinda funny then that amd was eating their lunch (and yeah i do remember that, the athlon XPs) with a more traditional cpu design, while almost remaining competitive clock-for-clock lol. i suppose that's what gave rise to merom (and i did have one of the original pentium Ms — clocked at 1.5 ghz it blew my previous, much thicker and hotter 2.4ghz p4 laptop, straight outta the water lol)
|
# ? Apr 21, 2024 15:47 |
|
Wild EEPROM posted:the first retail 4ghz intel chip wasnt netburst even though it was trivial to overclock well past that i had a 4770k at the time (still do even lol ) and ran it at 4.2 ghz pretty much since i got it. afaik it's still working fine today, although it's been powered down for almost a year now and is in storage
|
# ? Apr 21, 2024 15:50 |
|
Pendragon posted:remember how the first pentium 4s used Rambus RAM? remember those stupid spacer rambus rdimms that you had to use in what would otherwise be empty slots and were almost as expensive as an actual, populated rdimm? lol, lmao so glad they eventually supported ddr lol
|
# ? Apr 21, 2024 15:52 |
|
its very funny that intel still takes the "throw more power at the chip until we have the biggest number" approach with their top level stuff in a desperate attempt to stay relevant anyway, the intel "snake oil" presentation that leaked to the press was hilarious
|
# ? Apr 21, 2024 16:21 |
|
Intel thought they were going to get a 1000x speed boost from Netburst 10x from clock speed increases 10x from process shrinks 10x from multi-processing
|
# ? Apr 21, 2024 16:38 |
|
njsykora posted:its very funny that intel still takes the "throw more power at the chip until we have the biggest number" approach with their top level stuff in a desperate attempt to stay relevant lol i forgot about this tbh they are right, amd's numbering scheme is bullshit but so is intel's, so they're being extremely hypocritical here of course
|
# ? Apr 21, 2024 16:48 |
|
I gave up on trying to interpret cpu model numbers a long time ago You just have to look up the actual specs and probably benchmarks
|
# ? Apr 21, 2024 17:58 |
|
mystes posted:When I decide I need better cooling for my cpu I will remove massive wad of dust from the fan I opened my case for the first time in probably 3 years recently to change the GPU and was amazed that it was almost completely dust free thanks to the integrated filters and I guess the flow direction. ah the days of accumulating big balls of dust and having GPUs that just vaguely moved air around and having zero outtake fans
|
# ? Apr 21, 2024 18:30 |
|
mystes posted:I gave up on trying to interpret cpu model numbers a long time ago I just pick whatever processor someone's build guide suggests now. And the next computer I get is probably going to be prebuilt, I'm done with doing this myself now.
|
# ? Apr 21, 2024 22:29 |
|
Volmarias posted:I just pick whatever processor someone's build guide suggests now. And the next computer I get is probably going to be prebuilt, I'm done with doing this myself now. That was my plan but turns out the only prebuilts or even barebones on the market were office shitboxes or overpriced 'Pro gamer' stuff so I had to build my own like a caveman. Normal people buy laptops now I guess.
|
# ? Apr 21, 2024 23:00 |
|
during the start of the parts shortage in april 2020 I bought microcenter's in-house pre-built powerspec and it's been okay. not sure if I'd do it again, but it was the only gpu source available and my old system fried. at least the case is a sensible Lian-li and the cooling systems lights could be turned off the parts were all good, too, but not the brands I'd pick. good for the moment, though, since individual stock was spotty everywhere of gouging Agile Vector fucked around with this message at 23:12 on Apr 21, 2024 |
# ? Apr 21, 2024 23:07 |
|
Before I bought my current desktop in 2017 I thought I probably wouldn't bother to build a computer again but at that time it ended up working out much better to buy a mobo/cpu bundle from microcenter than buying a prebuilt one No idea what the situation is like now, although I probably should upgrade again at some point.
|
# ? Apr 21, 2024 23:33 |
|
|
# ? May 14, 2024 08:57 |
|
Crazy Achmed posted:yeah, but the dual cd format of c&c let both people play singleplayer as well as multi, which was extremely economical that is p cool
|
# ? Apr 22, 2024 02:03 |