|
"I guess we'll just have to ACQUIRE the next revolution in CPU design. Georgie, buy me everyone." "Who?" "EV-ERY-ONNNNNE!"
|
# ? Jun 11, 2021 04:14 |
|
|
# ? May 18, 2024 04:19 |
|
Good way to kill or marginalize something is to have intel buy itquote:Naveen Rao, who cofounded deep-learning chip startup Nervana, which was bought and later axed by Intel, warned SiFive could be a casualty of another war between Intel and Nvidia. Intel acquired Nervana to take on Nvidia's AI accelerators, and pulled the plug on that line of attack in early 2020. What could happen to SiFive if it's bought just to be a nuisance for a combined Nvidia-Arm? True but lol get hosed Naveen WhyteRyce fucked around with this message at 04:38 on Jun 11, 2021 |
# ? Jun 11, 2021 04:31 |
|
https://www.youtube.com/watch?v=YCRoo0tqJL4
|
# ? Jun 11, 2021 04:35 |
|
Wow that SiFive, definitely seems like a “capture and kill” thing.
|
# ? Jun 11, 2021 07:11 |
|
drat, too bad for those desktop skeptics who were hoping for a 8+0 S1 variant A fitting end for those who wanted to turn RISC-V into just another IP licensor. On the bright side, at least it isn't a China-sponsored group looking to buy them...?
|
# ? Jun 11, 2021 10:42 |
|
I REALLY want to see what Alder Lake-M looks like, that may be my next thin-and-light, if AMD can't put together an APU with RDNA cores before then.
|
# ? Jun 12, 2021 01:58 |
|
A friend asked for my help for running 3200C16 RAM on his ancient 7700K + Asus Z270 board, and the first thing I checked and found to none of my surprise is the mobo XMP pumping a ridiculous 1.33V into the VccIO and SA. Disabled XMP and dropped both to 1.1V without issues. I can imagine overselling RAM OCs are definitely going to be fun on DDR5
|
# ? Jun 14, 2021 04:07 |
|
Nomyth posted:drat, too bad for those desktop skeptics who were hoping for a 8+0 S1 variant lol at anyone who doesn't want this I'm sure the Intel+Microsoft combo will figure out how to screw it up, but having small cores is a great thing on M1 Macs. quote:A fitting end for those who wanted to turn RISC-V into just another IP licensor. On the bright side, at least it isn't a China-sponsored group looking to buy them...? I am struggling to understand the nature of your complaint. RISC-V is an open architecture. SiFive is one of many implementors. Regardless of how they choose to sell their work, there is nothing preventing other organizations from designing their own RISC-V cores, and in fact many are. It's also super weird that you apparently think it's bad for SiFive to pursue a business model where they license IP. What's wrong with that? also lol at the china paranoia
|
# ? Jun 15, 2021 23:38 |
|
BobHoward posted:lol at anyone who doesn't want this Part of me wants the 8+0 too, I agree with you on both sides. quote:I am struggling to understand the nature of your complaint. RISC-V is an open architecture. SiFive is one of many implementors. Regardless of how they choose to sell their work, there is nothing preventing other organizations from designing their own RISC-V cores, and in fact many are. It's...almost a personal thing. I got a little misled by the RISC-V foundation fomenting the idea that open architecture = anyone in their garage could start producing designs on their own with relatively little training, like with FOSS software. That turned out to be not the case, since you know...it takes a ridiculous amount of investment and manpower to produce and then debug hardware designs in general And part of the reason I'm even here in the first place typing this at you is because so much of CCP policy in regulating business has been, and still is, hilariously misguided
|
# ? Jun 16, 2021 02:13 |
|
Nomyth posted:It's...almost a personal thing. I got a little misled by the RISC-V foundation fomenting the idea that open architecture = anyone in their garage could start producing designs on their own with relatively little training, like with FOSS software. That turned out to be not the case, since you know...it takes a ridiculous amount of investment and manpower to produce and then debug hardware designs in general wait........what?
|
# ? Jun 16, 2021 02:27 |
|
can't believe the Communist Party of China is making it impossible for a single person to fab their own superscalar out-of-order multi-core RISC CPUs in their garage, smdh e: Sarcasm aside, did anyone ever do an OP for a non-x86 CPU architectures thread? Or is there even really enough interest in having one?
|
# ? Jun 16, 2021 07:53 |
|
I recently changed my cooler from the stock Intel fan to a DeepCool Gammaxx 400 Pro, and while I'm still utterly terrified of breaking a pin in the socket (which to be fair didn't really come into play here since I never removed the CPU from the socket), it was also a relief that the CPU retention bracket meant that I didn't have to worry about the CPU getting torn off the socket because of being glued to the cooler like with AMD. If the rumors are true, the next AMD socket is going to also be LGA anyway, so might as well pick up some experience.
|
# ? Jun 16, 2021 07:54 |
|
gradenko_2000 posted:I recently changed my cooler from the stock Intel fan to a DeepCool Gammaxx 400 Pro, and while I'm still utterly terrified of breaking a pin in the socket (which to be fair didn't really come into play here since I never removed the CPU from the socket), it was also a relief that the CPU retention bracket meant that I didn't have to worry about the CPU getting torn off the socket because of being glued to the cooler like with AMD. not gonna lie ryzen being PGA is partially why I haven't made the switch yet, do not wanna deal with that possibility
|
# ? Jun 16, 2021 09:08 |
|
It's not a very good reason, really.
|
# ? Jun 16, 2021 09:09 |
|
I suppose the advantage of being able to repair bent pins on the CPU has become less and less of an issue since I've stopped buying old second-hand parts from Aliexpress that come in a rudimentary plastic case entombed in six layers of plastic wrap, but I did get to resurrect two CPUs that way
|
# ? Jun 16, 2021 09:13 |
|
Honestly I ended up sticking with Intel instead of going AMD on my last build by basically complete accident (I was gonna migrate my old CPU opted against it at the last second… mostly because I bought the wrong motherboard when I was rushing and then just said gently caress it and gave my brother my old computer). I have really, really liked the 10900KF though. Good thermals, OC’d like a dream. I was disappointed the 11900 was basically a step down in every way, I was hoping Intel would still be good in the “more money than sense” market.
|
# ? Jun 16, 2021 10:09 |
|
LRADIKAL posted:It's not a very good reason, really. oof, but here goes, it took me a couple tries to get just how gently the chip needs to be dropped into the socket bed (just touching slowly). I've never messed up a motherboard beyond functionality - but I have damaged at least 1 pin visually on let's say 10% of my uninstall cycles as an amateur. I'm very ginger especially on consumer lga installations - I think the HEDT sockets (2011) were easier. The electrical (signaling/power) benefits of a consumer LGA socket are undeniable, and I don't think you'd get that on a PGA socket with the necessary density. It's the tradeoff of having pcie4 and ddr5 and poo poo. AMD has to spec AM5 for the next 5 years. Paul MaudDib fucked around with this message at 10:47 on Jun 16, 2021 |
# ? Jun 16, 2021 10:14 |
|
gradenko_2000 posted:being able to repair bent pins I feel like LGA is the way to go for reusability though. Simply by nature of being a planar surface with a bunch of stuff sticking out on top, the motherboard pin array is well protected. On the flip side, we don't need to worry about protecting CPU pins or repairing CPU pins when there are no CPU pins.
|
# ? Jun 16, 2021 13:01 |
|
they both suck, bring back slot 1/slot a
|
# ? Jun 16, 2021 19:10 |
|
Nomyth posted:It's...almost a personal thing. I got a little misled by the RISC-V foundation fomenting the idea that open architecture = anyone in their garage could start producing designs on their own with relatively little training, like with FOSS software. that's... not how FOSS works either?
|
# ? Jun 16, 2021 19:45 |
|
The RV-Foundation committees aren't open, are they? They just publish the minutes afterwards I think. So kinda of an 'AnDroID iS oPEn sOUrCe' situation. Related post: https://lore.kernel.org/linux-riscv/mhng-d64da1be-bacd-4885-aaf2-fea3c763418c@palmerdabbelt-glaptop/
|
# ? Jun 16, 2021 19:55 |
|
Nomyth posted:Part of me wants the 8+0 too, I agree with you on both sides. I mean, yes, it does cost a lot to tape out a chip. Not sure why you're bitter at SiFive over it; it's just reality. More to the point, they aren't the RISC-V foundation. And Dr. Fishopolis has an excellent point about that not being how anything actually works in FOSS software either. The C language spec is open and relatively small, but it takes millions of hours of effort (much of which is commercially sponsored) to create compilers as good as GCC and Clang. You don't see a thousand new compilers bloom all the time. Actually... you probably do, but you just don't hear about them because they're toy projects made because someone wanted to explore writing their own compiler. Since Clang and gcc exist and are a free download away, nobody pays much attention to the others. And that's probably true of RISC-V too. It's not possible for amateurs to create a cutting edge CPU, but a basic one made to scratch an itch? Absolutely yes. I have personal experience. My alma mater had students choose two of several possible design project courses in the final year of our undergrad CE degree program. One of my picks meant that over one semester, my lab partner and I designed a classic 32-bit pipelined RISC (iirc using a subset of the MIPS ISA, aka RISC-V's predecessor) and implemented it in a FPGA board. You wouldn't want to use that CPU. It didn't have a MMU, FPU, or caches. But it ran test programs, and with a little more experience than we had going in, it could've been much better. (To give you an idea how green we were, the course was also our introduction to designing logic with a HDL.) So, I bet there have been several weekend warrior FPGA RISC-V projects. They just don't get much press because there's not much reason for anyone to care. (both Altera-cough-Intel and Xilinx-cough-AMD ship soft CPU cores that can boot Linux, you get them for as close to free as anything gets in the FPGA world, and they're better than anything an amateur can easily create)
|
# ? Jun 16, 2021 20:26 |
|
No need to read further into my poorly thought-out posts. I'm just mad that I'm not Jim Keller, and my attention span + ability to post coherently are rotting away with the rest of my brainBobHoward posted:So, I bet there have been several weekend warrior FPGA RISC-V projects. They just don't get much press because there's not much reason for anyone to care. (both Altera-cough-Intel and Xilinx-cough-AMD ship soft CPU cores that can boot Linux, you get them for as close to free as anything gets in the FPGA world, and they're better than anything an amateur can easily create) Yep I've used the Altera/Intel FPGA one myself before (Nios II). I'm about 10+ years too late to be in the work of creating the tools that everyone else would leverage for their problem-solving devices
|
# ? Jun 16, 2021 21:39 |
|
Paul MaudDib posted:oof, but here goes, it took me a couple tries to get just how gently the chip needs to be dropped into the socket bed (just touching slowly). I've never messed up a motherboard beyond functionality - but I have damaged at least 1 pin visually on let's say 10% of my uninstall cycles as an amateur. motherboards tend to be cheaper than CPUs (unless you go ham with one of those super expensive ones i guess) so if i had to break one or the other i'd rather break the motherboard
|
# ? Jun 17, 2021 00:30 |
|
Cygni posted:they both suck, bring back slot 1/slot a gently caress yeah. I never broke any of these ever. SMASH IN, DONE! On the other hand I make loving drat sure i never drop a LGA anything into a socket, I clean my fingers off, hold the cpu very carefully and then gingerly float it into the socket followed by ramming the LGA bracket down on top of the CPU. So far no gently caress ups.
|
# ? Jun 17, 2021 01:26 |
|
Every day, I wake up and I open palm slap a Pentium II into the slot...
|
# ? Jun 17, 2021 01:40 |
|
Cygni posted:they both suck, bring back slot 1/slot a I can’t remember how I got it but I had an adapter board for slot a that let you put a non slot PIII chip in it. I had it in some hand me down IBM desktop machine. Alongside the matrox gpu, that was my first start with pc gaming.
|
# ? Jun 17, 2021 12:13 |
|
track day bro! posted:I can’t remember how I got it but I had an adapter board for slot a that let you put a non slot PIII chip in it. I had it in some hand me down IBM desktop machine. Alongside the matrox gpu, that was my first start with pc gaming. Pretty sure this was a "Slocket" https://en.wikipedia.org/wiki/Slotket
|
# ? Jun 17, 2021 12:19 |
|
Yeah it was one of those, I remember it having a fan that went on top of the cpu I think? But I definitely didn't use any thermal paste, either I didn't know or it wasn't a thing?
|
# ? Jun 17, 2021 14:33 |
|
BobHoward posted:So, I bet there have been several weekend warrior FPGA RISC-V projects. They just don't get much press because there's not much reason for anyone to care. (both Altera-cough-Intel and Xilinx-cough-AMD ship soft CPU cores that can boot Linux, you get them for as close to free as anything gets in the FPGA world, and they're better than anything an amateur can easily create) A couple years back I tried and was generally disappointed with the OTS risc-v stuff, there were containers to bundle up the tools. Fabscalar might have been around then? It would've been fine if I was a grad student with infinite tolerance for juggling options around but I was quickly frustrated at how not-turnkey it was despite seeming so close.
|
# ? Jun 17, 2021 19:09 |
|
JawnV6 posted:A couple years back I tried and was generally disappointed with the OTS risc-v stuff, there were containers to bundle up the tools. Fabscalar might have been around then? It would've been fine if I was a grad student with infinite tolerance for juggling options around but I was quickly frustrated at how not-turnkey it was despite seeming so close. That was the exact impression my co-workers got when we tried the same, during my internship.
|
# ? Jun 18, 2021 01:52 |
|
Kazinsal posted:e: Sarcasm aside, did anyone ever do an OP for a non-x86 CPU architectures thread? Or is there even really enough interest in having one? As someone who has been lusting after one of Raptor Computing's POWER9 setups for years, that would be welcome. Anybody wants to start talking about the ghosts of SPARC and MIPS, SMT4 and SMT8, and what server-targeted ARM is like, I'd be there.
|
# ? Jun 18, 2021 03:06 |
|
Hasturtium posted:As someone who has been lusting after one of Raptor Computing's POWER9 setups for years, that would be welcome. Anybody wants to start talking about the ghosts of SPARC and MIPS, SMT4 and SMT8, and what server-targeted ARM is like, I'd be there. Awesome. I started working on an OP a while ago but lost what I had in a power blip so I’ll probably start a new thread this evening and write about the architectures I know about and let other people contribute primers for the ones I don’t.
|
# ? Jun 18, 2021 03:13 |
|
Vanagoon posted:Pretty sure this was a "Slocket" Slocket! Hell yeah, the worst of both worlds!
|
# ? Jun 18, 2021 03:24 |
|
Kazinsal posted:Awesome. I started working on an OP a while ago but lost what I had in a power blip so I’ll probably start a new thread this evening and write about the architectures I know about and let other people contribute primers for the ones I don’t. Thank you, link it here. I'd love to find a happy medium between "set up a cluster of Raspberry Pi's" and "drop three grand for a POWER setup" that doesn't involve raiding eBay for a closeout server from over a decade ago, or somebody's microATX Amiga non-starter.
|
# ? Jun 18, 2021 03:26 |
|
Hasturtium posted:Thank you, link it here. I'd love to find a happy medium between "set up a cluster of Raspberry Pi's" and "drop three grand for a POWER setup" that doesn't involve raiding eBay for a closeout server from over a decade ago, or somebody's microATX Amiga non-starter. Awesome! Here goes: https://forums.somethingawful.com/showthread.php?threadid=3970988
|
# ? Jun 18, 2021 06:06 |
|
Does this mean we're threatening to merge the AMD and Intel threads but not actually do it again?
|
# ? Jun 18, 2021 12:17 |
|
Thanks! I chipped in a quick blurb about Power to get the ball rolling.
|
# ? Jun 18, 2021 12:38 |
|
SwissArmyDruid posted:Does this mean we're threatening to merge the AMD and Intel threads but not actually do it again? I honestly wouldn't mind seperate AMD/Intel/Nvidia tech support and buying advice threads and a merged pc hardware news thread
|
# ? Jun 18, 2021 12:42 |
|
|
# ? May 18, 2024 04:19 |
|
The GPU thread moves way too fast to merge with either this or the AMD thread.
|
# ? Jun 18, 2021 20:23 |