|
FaustianQ posted:Why not an ARM solution? Or is legacy code holding them back? God NASA needs an actual loving budget. Because the 486 can survive radiation in space and space programmers like to base things around clockrate like old dos games.
|
#
?
Mar 1, 2017 16:18
|
|
|
|
| # ? May 18, 2024 04:07 |
|
|
Hits turbo button on space shuttle
|
|
#
?
Mar 1, 2017 16:26
|
|
|
My understanding was that the rad hardened processors of choice have usually not been 386 or 486 but rather POWER based architectures, stuff like the RAD6000 and RAD750.
|
|
#
?
Mar 1, 2017 16:29
|
|
|
Truga posted:It's this. This isn't even a time constraint on certifying things. A random gamma ray (and there are a lot in space) hitting the big 386 chip doesn't really have much chances of displacing something important. A high energy neutron hitting something in your modern <30nm arch is going to wreck it hard. 386 and 486 are perfect for that use, just due to the sheer size of each element. My uncle used to work at NASA and one of his jobs was to take laptops up to Indiana University in Bloomington and put them in the particle accelerator to see how long they'd last (answer: not long). He never did get me into the sims but he did give me his Silver Snoopy award a couple years ago.
|
|
#
?
Mar 1, 2017 17:01
|
|
|
Rastor posted:My understanding was that the rad hardened processors of choice have usually not been 386 or 486 but rather POWER based architectures, stuff like the RAD6000 and RAD750. You see those more often on unmanned craft, like space probes or rovers. For instance, all the Mars rovers that currently active are based on radiation hardened POWER CPUs. There's actually a paper from NASA in 1991 going over the pros and cons of whether the Space Station Freedom (predeccesor project to the ISS, and which formed the basis for most of the modules that the US would put up for the ISS) design should use 386s or 486s: https://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19910016373.pdf
|
|
#
?
Mar 1, 2017 17:13
|
|
|
me: "There's no way AMD could cherry-pick settings to pad their Cinebench numbers, Cinebench doesn't have any settings!" https://twitter.com/CPCHardware/status/836967901399744513 The press/preview BIOS has a special profile tuned specifically for Cinebench 
|
|
#
?
Mar 1, 2017 17:28
|
|
|
Deuce posted:I want to see Ryzen benchmarks compared to the ISS's 486 processors. AMD processors in space: 0 Intel processors in space: more than 0 fishmech posted:And it's worth remembering that the laptops and such which they use for the research are just normal modern Thinkpads most of the time. They've got about 50 or so thinkpads across the station. Even cooler the ISS has a cargo hold with thinkpads that's depressurized so if you need a new laptop you can go pick one up and don't have to wait on earth to send you a new one. Also I hate ARM and the buzzword it's become. Yes it's in phones and everyone know it's a thing now and 
|
|
#
?
Mar 1, 2017 17:32
|
|
|
repiv posted:me: "There's no way AMD could cherry-pick settings to pad their Cinebench numbers, Cinebench doesn't have any settings!" Wouldn't the numbers published by Intel have similar tweaks? If some technician changed all the BIOS settings by hand before running the bench, would that make it better? It is a little gauche, but I'd expect, for example, a PR benchmark of a database produced by the DB's creator to have been tweaked for max performance.
|
|
#
?
Mar 1, 2017 17:34
|
|
|
wccf has a bunch of new leaked slides http://wccftech.com/amd-ryzen-architecture-slide-review-leak/
|
|
#
?
Mar 1, 2017 17:38
|
|
|
taqueso posted:Wouldn't the numbers published by Intel have similar tweaks? If some technician changed all the BIOS settings by hand before running the bench, would that make it better? It is a little gauche, but I'd expect, for example, a PR benchmark of a database produced by the DB's creator to have been tweaked for max performance. Yes, but it's in poor taste to get caught.
|
|
#
?
Mar 1, 2017 17:39
|
|
|
Boiled Water posted:AMD processors in space: 0 Ok ok but what if we used TriCore???
|
|
#
?
Mar 1, 2017 17:40
|
|
|
They are, or were, working on a rad-hardened multi-core ARM processor. And the older Intel processors running spacecraft and satellites have mostly been supplanted by PowerPC chips. I don't think anyone cares what architecture these things use or what clock speed they run. Anyway When are we going to get some real Ryzen reviews?
|
|
#
?
Mar 1, 2017 17:41
|
|
|
Voyager I posted:Yes, but it's in poor taste to get caught. cinebunch.exe
|
|
#
?
Mar 1, 2017 17:43
|
|
|
FuturePastNow posted:When are we going to get some real Ryzen reviews? Tomorrow.
|
|
#
?
Mar 1, 2017 17:44
|
|
|
FuturePastNow posted:When are we going to get some real Ryzen reviews?
|
|
#
?
Mar 1, 2017 17:46
|
|
|
The Space Shuttle ran on Intel 8086 chips lol. They event had to go to eBay to pick up some for the last few space shuttle launches I heard back in the day. Also the New Horizons probe uses a modified Playstation 1 CPU.
|
|
#
?
Mar 1, 2017 18:15
|
|
|
EdEddnEddy posted:The Space Shuttle ran on Intel 8086 chips lol. They event had to go to eBay to pick up some for the last few space shuttle launches I heard back in the day. Nah, by the time eBay was around the space shuttles were being upgraded to 386-based systems, which they then retained for the rest of the time they were in operation. Also New Horizons is only using a modified Playstation 1 CPU in the same way that the RAD750 in the latest Mars Rover is a modified GameCube CPU - New Horizons uses a modified MIPS R3000 core, just as the PlayStation uses a modification of the R3000 core, but they're entirely different modifications with little relation to each other. Notably, the variant used to create the processor in the New Horizons probe derives from refinements made for embedded use and adds very few instructions, while the PSX CPU had more radical changes. Similarly, the CPU in the GameCube (and from that the Wii and Wii U) is derived from a modified PowerPC 750 (sold in macs as a "PowerPC G3" CPU) and so is the RAD750 on the latest mars rover. But the GameCube version of it adds a lot of instructions and subtracts a few, and derives from one revision of the series ultimately, while the RAD750 is derived from a different later revision with very few instuction changes, instead with the majority of effort spent on radiation hardeneing and improved processes so that it uses less power.
|
|
#
?
Mar 1, 2017 18:33
|
|
|
Well it was a while ago, but I remember reading about It when it was news.* I could have sworn it was an Engadget article or something but alas, time has passed as has the SS. *Whoever is the editor who allowed no proper capitalization should be shot.
|
|
#
?
Mar 1, 2017 18:50
|
|
|
EdEddnEddy posted:Well it was a while ago, but I remember reading about It when it was news.* I could have sworn it was an Engadget article or something but alas, time has passed as has the SS. That's covering a different issue: NASA needed 8086s for on-the-ground equipment used to test various components of the shuttles and their rocket boosters between launches and in the stages leading up to launches. After all, they couldn't have used regular consumer or even most military 8086s in the space shuttle, as they wouldn't be properly radiation hardened. But the ground equipment could use any 8086 just fine. The first shuttle to have the main computers upgraded from 8086 to 80386 was the Atlantis in 1997, and the remaining 4 shuttles had the 386s installed one shuttle at a time, and upgrading a single shuttle each year. The intention of that staggered schedule, which took until 2001, was that there would be an old system shuttle ready to fly at any time in case the 386s turned out to have issues.
|
|
#
?
Mar 1, 2017 19:16
|
|
|
fishmech posted:That's covering a different issue: NASA needed 8086s for on-the-ground equipment used to test various components of the shuttles and their rocket boosters between launches and in the stages leading up to launches. After all, they couldn't have used regular consumer or even most military 8086s in the space shuttle, as they wouldn't be properly radiation hardened. But the ground equipment could use any 8086 just fine. Was that around the time they went to a more Glass cockpit as well? It's crazy some of the early pics of the SS cockpit vs the later stuff. And as bad/crazy as the SS was for NASA, it did some amazing things and really was impressive even if always dangerous in one form or another. The one thing I regret is never seeing a Space Shuttle launch in person.
|
|
#
?
Mar 1, 2017 19:22
|
|
|
I think that the SpaceX rocket and Dragon run on ARM. Not entirely sure though. We do know that they certainly don't use hardened processors for Dragon. In other news....Kyle Bennet on an AMD event? WTF is happening?  https://www.youtube.com/watch?v=ZVKDNeyfpAo&t=4965s
|
|
#
?
Mar 1, 2017 19:26
|
|
|
Kyle is a lot handsomer than I thought he'd be 
|
|
#
?
Mar 1, 2017 19:32
|
|
|
EdEddnEddy posted:Was that around the time they went to a more Glass cockpit as well? It's crazy some of the early pics of the SS cockpit vs the later stuff. Yeah a major impetus behind moving to the 386 systems was that the extra processing power was needed to drive the "glass cockpit" sort of displays.
|
|
#
?
Mar 1, 2017 19:41
|
|
|
did I miss the part where AMD announced a partnership with SpaceX to shoot a Ryzen CPU around the moon or where's all that space chat coming from?  A full review by an a persian (iranian?) site is up here: http://translate.google.com/transla...n&langpair=auto  Allegedly this is a 1700X sample with a prerelease BIOS stuck at 3.4 Ghz with 2133 Mhz RAM. No Turbo. Here's a german guy trying to delid Ryzen CPUs, casually mentioning that he already killed two in the process. They're soldered by the way. It's also pretty interesting that it looks like the 8C/16T models have two separate dies closely next to each other. https://www.youtube.com/watch?v=lOZbK3tP7EU eames fucked around with this message at 20:21 on Mar 1, 2017 |
|
#
?
Mar 1, 2017 19:59
|
|
|
Makes sense that a large 8core part is soldered, interesting finally seeing a configuration different from Xeon style square dies, too. The two die thing matches up with those architecture slides, it looks like their goal was to build chips out of multiple 4 core packs, sort of like current GPUs and shader modules.
|
|
#
?
Mar 1, 2017 20:26
|
|
|
Its not 2 seperate dies I think. That looks like that due to how they laid down the metal TIM in 2 different pads side by side. Also MCM's are real expensive.
|
|
#
?
Mar 1, 2017 21:02
|
|
|
Oh look, a new chip! Time for a whole sparkly new system.. with a 1080ti. Ryzen and Vega bundles on launch would have been awesome. 
GRINDCORE MEGGIDO fucked around with this message at 21:11 on Mar 1, 2017 |
|
#
?
Mar 1, 2017 21:04
|
|
|
Voyager I posted:it's in poor taste "marketing" Dante80 posted:I think that the SpaceX rocket and Dragon run on ARM. Not entirely sure though. We do know that they certainly don't use hardened processors for Dragon. Seems like a bad idea to trust software division on a guidance system but I guess I'm not a very good distruptor
|
|
#
?
Mar 1, 2017 21:06
|
|
|
eames posted:did I miss the part where AMD announced a partnership with SpaceX to shoot a Ryzen CPU around the moon or where's all that space chat coming from?  Looks like you may want to change how you apply thermal paste if you normally do a grain of rice. This looks more like a job for a line across the middle. (And you all thought I was kidding about wanting to see a delidded Ryzen chip so I knew how to properly be pasting it.) Also, apparently the entire slide deck for tomorrow's release has been "leaked". RIP whoever "Leynar Khayrullin" is, if that's even a real person. (I bet it's not.) SwissArmyDruid fucked around with this message at 21:34 on Mar 1, 2017 |
|
#
?
Mar 1, 2017 21:23
|
|
|
My original concern about moving to something other than old 486/386 processors to something like ARM wasn't about performance but rather how much they could drive down the thermals. My understanding of x86 is it's really hard to push down into the milliwatt range, especially such old designs. Interesting that they use POWER designs as well.
|
|
#
?
Mar 1, 2017 21:32
|
|
|
https://drive.google.com/file/d/0By7_TCzoyL9oejlBV1R0WVdKMkU/view
|
|
#
?
Mar 1, 2017 21:38
|
|
|
  ....still think AMD doing this is dumb. It'll go into my spare parts drawer for when I need something to hold me over while waiting for parts to repair/replace my AIO cooler, and I'll never even think about it otherwise, but this just eats into margin, and I *still* think AMD is leaving money on the table.
|
|
#
?
Mar 1, 2017 21:41
|
|
|
 Random numbers!
|
|
#
?
Mar 1, 2017 21:42
|
|
|
sounds like reviewers are going to be working late tonight https://twitter.com/GamersNexus/status/837026790526881797
|
|
#
?
Mar 1, 2017 21:45
|
|
|
repiv posted:sounds like reviewers are going to be up late tonight They found if you turn off the cheating bios mode it really changes the results.
|
|
#
?
Mar 1, 2017 21:46
|
|
|
FaustianQ posted:My original concern about moving to something other than old 486/386 processors to something like ARM wasn't about performance but rather how much they could drive down the thermals. My understanding of x86 is it's really hard to push down into the milliwatt range, especially such old designs. That's not really all that useful in these applications if it can't provide the appropriate performance, especially since radiation-hardening procedures can easily result in higher power draw as part of ensuring robustness. For perspective, the RAD750 hardened PowerPC chip draws up to 5 watts of power. Normal embedded versions of the PowerPC 750 it was based on drew around 4.5-4.9 watts while the original CPU drew 7 watts when first launched in 1997 (the RAD750 design variant came out in 2001). There happen to be a few ARM based CPUs out there that are radiation hardened right now, but they tend to be pretty slow compared to existing radiation-hardened CPUs. RAD750s installed in functioning spacecraft run at up to 200 MHz at the moment, while most of the ARM Cortex based chips available from companies like Vorago are running at speeds like 50 MHz or 75 MHz, and with less usable computation power per cycle as well - albeit those chips do only use 0.25 - 0.5 watts at full load. (For what it's worth those ARM chips are comparable in performance to the modified 20 MHz 386s used for main system control in the ISS and space shuttle, but that's not considered suitable for new projects, particularly when image processing is needed) NASA and the ESA are currently working on projects with industry to develop radiation-hardened ARM CPUs fast enough to replace existing CPUs in spacecraft/probes/rovers, but they do not expect the project to produce workable product until 2020 at the earliest, which means waiting until like 2025 for craft using those CPUs to actually get designed and launched.
|
|
#
?
Mar 1, 2017 21:59
|
|
|
repiv posted:sounds like reviewers are going to be working late tonight Reading between the lines, you need to test in a specific way to make AMD look good instead of just running benchmarks or programs like normal people. loving AMD.
|
|
#
?
Mar 1, 2017 22:02
|
|
|
redeyes posted:Reading between the lines, you need to test in a specific way to make AMD look good instead of just running benchmarks or programs like normal people. loving AMD. It's a really dumb move, AMD clearly has (soon had) the whole enthusiast community, including reviewers, rooting for them as the underdog that may not deliver 100% of the gaming performance but at least tries to change the desktop CPU landscape for the better. Now it increasingly looks like throwing it all away with BIOS "benchmark modes", strict last minute reviewing guides and whatever else. It's not like those additional few percent would have changed the buying decision of many people. The truth will come out either way.
|
|
#
?
Mar 1, 2017 22:10
|
|
|
On the other hand, there are rumors about Ryzen having a poo poo-ton of errata for it, so they could just be pushing out bug fixes and not wanting to have the chips forever marred by bad benchmarks from the initial release getting listed in charts and compared to other stuff, regardless of what the reality of the situation is, which is something that's bitten them in the rear end before.
|
|
#
?
Mar 1, 2017 22:14
|
|
|
|
| # ? May 18, 2024 04:07 |
|
|
for the more casual followers of this sort of stuff, what does "errata" refer to in a CPU?
|
|
#
?
Mar 1, 2017 22:17
|
|
