|
Laslow posted:Jot this little nugget down, wall street: It's obviously Itanium 3. It probably wouldn't be a bad idea at this point. A huge percentage of corporate server farms are busy running .Net or Java stuff at this point. If Qualcomm is banking on people rewriting or developing middleware to allow generic software to run on ARM chips, why not a new ISA?
|
# ? Apr 29, 2018 15:15 |
|
|
# ? Apr 27, 2024 17:31 |
|
Xae posted:It probably wouldn't be a bad idea at this point. If you were to design an architecture to be unsuitable for most business workloads you would have to try very hard to be worse than Itanium is.
|
# ? Apr 29, 2018 16:09 |
|
movax posted:but I’m wondering what the next IP core to get laid down is If it were up to me, fpga. Or, if that seems to complex for the benefit, several general-purpose compression/decompression cores. Like hardware zlib. A big part of the reason game load times aren't better between SATA and NVMe is because they're CPU-bound, and a big reason they're CPU-bound seems to be decompression. For decompression, 8-core CPUs can handle only like 600 MB/s of compressed data, while a hardware core should be able to handle the full 7 GB/s of a PCIe 4 NVMe SSD.
|
# ? Apr 29, 2018 18:08 |
|
Decompression is the least of an issue. I'd be surprised, if games use bog-standard compression algorithms these days. Audio and video is compressed with adequate codecs, gamedata is in god knows what shape. Latter usually has to be processed during load, that's what's limiting load times.
|
# ? Apr 29, 2018 18:45 |
|
Combat Pretzel posted:Decompression is the least of an issue. I'd be surprised, if games use bog-standard compression algorithms these days. Audio and video is compressed with adequate codecs, gamedata is in god knows what shape. Latter usually has to be processed during load, that's what's limiting load times. When I was figuring that bit out, I took a look at Skyrim. The file format they use generally consists of mostly zlib compressed data. Audio doesn't usually need up front decompression, so that's not relevant to load times, and textures were usually S3 compressed images that were then also zlib compressed in the pack file. (S3 and friends are designed to be stream-decompresed inside the gpu, so they're not actually very good compression overall and zlib can improve upon them for storage on disk.) Glancing around, this seems to be commonplace in modern games: more modern BC7 texture compression, then further compressed with standard zlib on disk. Taking decompression load off CPU would be a big help. With many existing games, you're right that there'd still be a lot of other processing going on, but that's something that's subject to optimization if the developers want to.
|
# ? Apr 29, 2018 20:37 |
|
movax posted:That was an interesting (and honestly entertaining) derail to read waiting in line for the Victoria Peak tram but lets say its run its course and cherish these page numbers forever. well, they'll likely dump some "ai core" -- read: fancy DSP -- and if they are smart they'll combine the uncore junk into one nice semi-programmable image/video/ai processor i actually think a proper hexagon style DSP is more likely esp after EMIB becomes mainstream and keller does his magic with SOC and fabric
|
# ? Apr 30, 2018 00:07 |
|
crazypenguin posted:When I was figuring that bit out, I took a look at Skyrim. The file format they use generally consists of mostly zlib compressed data. Audio doesn't usually need up front decompression, so that's not relevant to load times, and textures were usually S3 compressed images that were then also zlib compressed in the pack file. (S3 and friends are designed to be stream-decompresed inside the gpu, so they're not actually very good compression overall and zlib can improve upon them for storage on disk.) zlib is fundamentally not optimized for either speed or compression ratio but for ease of programming. for speed you can hit ram speed with something like LZ4 or zstd, or have a much better ratio with LZMA there's an entire company (basis) built around fast and not lovely compression b/c it matters for mobile games
|
# ? Apr 30, 2018 00:09 |
|
Malcolm XML posted:zlib is fundamentally not optimized for either speed or compression ratio but for ease of programming. for speed you can hit ram speed with something like LZ4 or zstd The first part is true, there's worthwhile stuff that's more than 4 times faster than zlib. But I'm really quite sure it's not near ram speed, not at all. Looking at the Xeon on https://quixdb.github.io/squash-benchmark and digging around it looks like you can get 4 GB/s out of 25 GB/s of theoretical memory bandwidth (and I'd guess about 20 GB/s of practical) with a fast LZ4 algorithm on all 4 cores. (Don't put too much stock in the 'copy' numbers on that page, I don't think they have really tried to make that actually efficient, it's just a point of comparison.) Anyway, I picked zlib mostly because (a) I know there are games that use it and (b) I know fpga cores exist that can eat it at GB/s, so it can definitely be accelerated in hardware.
|
# ? Apr 30, 2018 03:46 |
|
As was leaked earlier in that document that showed the new X399 compatibility, the bizarro Kaby Lake X CPUs (7640X and 7740X) are officially end of life. https://www.anandtech.com/show/12701/intel-announces-eol-plan-for-kaby-lakex-processors
|
# ? May 1, 2018 17:58 |
|
Cygni posted:As was leaked earlier in that document that showed the new X399 compatibility, the bizarro Kaby Lake X CPUs (7640X and 7740X) are officially end of life. They never had a chance. Poor little ones.
|
# ? May 1, 2018 19:41 |
|
Z390 + 8 core Coffee Lake Refresh (Whiskey Lake?) in Q3. https://www.anandtech.com/show/12711/german-it-distributor-publishes-amd-intel-roadmaps-z490-z390-8core-cfl-in-q4 e: actually looking at the super light, hard to see chart, it looks like Z390 in Q3 and 8 Core in Q4. Cygni fucked around with this message at 18:34 on May 3, 2018 |
# ? May 3, 2018 17:38 |
|
Cygni posted:Z390 + 8 core Coffee Lake Refresh (Whiskey Lake?) in Q3. Nice! I was waiting for an 8 core to upgrade from my 4770.
|
# ? May 3, 2018 17:45 |
|
14nm++++++++ launching in Q4 2022
|
# ? May 3, 2018 17:49 |
|
Looking forward to non-soldered 8-core Intels that go critical and burn a hole through your motherboard the moment you touch the multiplier.
|
# ? May 3, 2018 17:57 |
|
https://www.techspot.com/news/74447-eight-new-spectre-variants-affecting-intel-chips-discovered.html More and more spectre variants.
|
# ? May 3, 2018 17:59 |
|
repiv posted:Looking forward to non-soldered 8-core Intels that go critical and burn a hole through your motherboard the moment you touch the multiplier. Yeah, I'm already resigning myself to actually delidding this one.
|
# ? May 3, 2018 18:17 |
|
Cygni posted:Z390 + 8 core Coffee Lake Refresh (Whiskey Lake?) in Q3. An AMD 490 might explain why we only got the Hero tier of ASUS's Crosshair motherboards and not anything higher.
|
# ? May 3, 2018 18:38 |
|
repiv posted:Looking forward to non-soldered 8-core Intels that go critical and burn a hole through your motherboard the moment you touch the multiplier. but it'd reduce profit per unit by $1.23 so they won't
|
# ? May 3, 2018 21:15 |
|
Anime Schoolgirl posted:it's like they could use a excellent thermally conductive bonding material for this Intel is also able to cut a lot more corners by not soldering than just the TIM. You really see it in some of their 14nm processors. The TIM is also more reliable than solder in theory, but in reality we see almost no reports of damage to soldered IHS besides in the LN2 community which is super niche. Khorne fucked around with this message at 21:43 on May 3, 2018 |
# ? May 3, 2018 21:37 |
|
Is there any technical reason for them to not use liquid metal like every other delid-relid overclockers have been doing? Possible LM flowing to the wrong place? or is it not that good of a long term solution?
|
# ? May 3, 2018 21:47 |
|
Gamers Nexus had a video on that topic which amounted to "if they were going to go that far, they may as well solder". I think part of the issue is liquid metal could be shaken free of where it's supposed to be during transit.
|
# ? May 3, 2018 22:03 |
|
Xae posted:It probably wouldn't be a bad idea at this point. As mentioned Itanium is a turd that depended on magical compilers that were never going to exist. Also, AArch64 effectively is a new ISA. It's quite different from classic 32 bit ARM for performance reasons (no predication, no integrated barrel shifter). What do you propose having in a new ISA that's better? Even if you don't want to pay ARM royalties, there's RISC-V already.
|
# ? May 3, 2018 22:07 |
|
So no HEDT refresh in 2018? There were rumors of soldered Skylake-X a while ago.
|
# ? May 3, 2018 22:07 |
|
feedmegin posted:As mentioned Itanium is a turd that depended on magical compilers that were never going to exist. bring back i860
|
# ? May 3, 2018 22:26 |
|
Can't wait to see someone try and get that 8 core working on some rando z170 and the subsequent building fire
|
# ? May 3, 2018 23:18 |
|
Elon Musk confirmed that Jim Keller was going to work on a server architecture.quote:Romit Jitendra Shah - Nomura Instinet
|
# ? May 3, 2018 23:18 |
|
A radical new server architecture could be wicked cool but being intel it will be infinifabric something-or-other with slightly faster x86
|
# ? May 4, 2018 00:22 |
|
iAPX 432 rides again
|
# ? May 4, 2018 00:24 |
|
PCjr sidecar posted:iAPX 432 rides again Don't stop, I'm almost there!
|
# ? May 4, 2018 00:43 |
|
mewse posted:A radical new server architecture could be wicked cool but being intel it will be infinifabric something-or-other with slightly faster x86 Keller wants in on those Optane DIMMs
|
# ? May 4, 2018 00:46 |
|
HP The Machine only not a Meg Whitman boondoggle
|
# ? May 4, 2018 03:08 |
|
As long as the Pentium 4's horrifyingly awful netburst architecture never returns from the black pit of hell, then we good Intel. Netbust was fuckin' stupid.
|
# ? May 4, 2018 03:42 |
|
Vanagoon posted:As long as the Pentium 4's horrifyingly awful netburst architecture never returns from the black pit of hell, then we good Intel. maybe with emib and 10nm it will be good.
|
# ? May 4, 2018 03:44 |
|
coke posted:Is there any technical reason for them to not use liquid metal like every other delid-relid overclockers have been doing? Many reasons: 1. One gram of Conductonaut costs ~$16 retail. Intel would need *tons* of the poo poo. 2. Tons of liquid metal needs to be stored. No matter where you're assembling the chips, the local overseeing authorities are going to want to know that stew of heavy metals (Conductonaut evidently uses a mix of tin, gallium, and indium) is well-stored. 3. Gallium and indium are heavy metals. You could have the best safety and anti-exposure protocols in place, you're still going to be exposing workers to potential harm from long-term exposure. 4. When the end-user puts liquid metal TIM on their CPU, the CPU stays stationary and the end-user takes their time. When a factory worker is putting LM TIM on a CPU, they have quotas to reach, and said CPUs are then tossed around in handling and shipping. Honestly, what Intel *needs* to do is expand their existing and rather unnecessary-at-the-moment "overclockers insurance" to cover delidding so long as the end-user 1) uses a certified tool, and 2) submits a photograph to Intel showing no damage to the CPU from the process with the model/serial number clearly visible. *Intel* should then supply the end-user with a perfectly-machined shim to allow bare-core cooling. And if you think they'll ever do anything like that, boy do I have a bridge to sell you.
|
# ? May 4, 2018 04:34 |
|
Or Intel could use some other TIM that's still safe but better than whatever crap they currently use. Or they could solder the drat heatspreader on. There's many options here
|
# ? May 4, 2018 05:05 |
|
I was under the impression that the problem was the die-to-heatspreader clearance, rather than the TIM itself? Couldn’t Intel tighten up their clearances and/or use a different method of attaching the IHS to the chip package? When I delidded my 4790K, I just used whatever compound I had left over, but removing the old glue and sanding the edges of the IHS until it sat right on top of the chip itself gave me a 20 degree improvement under load.
|
# ? May 4, 2018 13:23 |
|
Or design a better shim so CPUs don't even need a heatspreader.
|
# ? May 4, 2018 16:56 |
|
lDDQD posted:Or design a better shim so CPUs don't even need a heatspreader. Or spec heatsinks to work on a naked die/package: https://www.kickstarter.com/projects/341758082/ncore-v1-naked-die-cooling-waterblock-designed-by
|
# ? May 4, 2018 17:04 |
|
Question that maybe some folks in this thread could answer. I've been looking at setting up a server on the cheap and came across this motherboard: https://www.newegg.com/Product/Product.aspx?Item=N82E16813182967&cm_re=dual_xeon_motherboard-_-13-182-967-_-Product It seems like a great deal, but I noticed it takes RDIMMs. Do I have to use registered memory with this board? Could I use unregistered?
|
# ? May 4, 2018 23:22 |
|
|
# ? Apr 27, 2024 17:31 |
|
VostokProgram posted:It seems like a great deal, but I noticed it takes RDIMMs. Do I have to use registered memory with this board? Nope, you can also use LRDIMMs! quote:Could I use unregistered? Nope! "Workstation" boards usually support UDIMMs and sometimes support RDIMMs. Server boards usually support RDIMMs and sometimes LRDIMMs. Look for an X99 board if you want to use UDIMMs.
|
# ? May 5, 2018 03:36 |