|
fishmech posted:(there's also hacky poo poo where you install with a compatible CPU then switch it out for one that lacks the NX bit, which would allow you to run Windows 10 on CPU from about 2001) It's amazing that works because on NX-equipped CPUs if you frob the NX bit on a page entry without having NX turned on you get a general protection fault. I guess that bit is just counted as "reserved, writes ignored" on pre-NX CPUs using PAE. I'd check for my own curiosity's sake but I don't have a sufficiently old copy of the Intel manuals... (ask me about my wealth of knowledge with regards to how fucky x86 and the x86 MMU is)
|
#
¿
Apr 20, 2017 19:23
|
|
|
|
| # ¿ Apr 19, 2024 21:18 |
|
|
Obsurveyor posted:ECC required for ZFS smells like bullshit and cargo cult behavior to me. 95% of what people say about ZFS is cargo cult behaviour. It doesn't need ECC and it doesn't need something ridiculous like two gigs of RAM per terabyte.
|
|
#
¿
May 5, 2017 22:10
|
|
|
Yeah, AMD distinctly doesn't suck in the CPU department anymore. GPUs are still a disillusioned sigh at the top end though.
|
|
#
¿
May 20, 2017 21:54
|
|
|
If anyone here is a super low level technical person and wants to know more about the conversion of x86 instructions to RISC micro-ops and associated pipeline/latency/clock/unit info, this is a really good PDF that goes into detail for pretty much every x86 microarchitecture since P5 and K7 (unfortunately no K6, which is where AMD introduced their RISC86 architecture): http://agner.org/optimize/instruction_tables.pdf
|
|
#
¿
May 28, 2017 07:32
|
|
|
I see 80-82 C with all cores maxed at 4.2 GHz/1.35V on my i7-3820. I really need a new CPU.
|
|
#
¿
Jun 4, 2017 02:44
|
|
|
SwissArmyDruid posted:There are two segments that AMD *needs* to make inroads in on: Server and mobile. That's where all the money is, and where turnover is most likely to happen. The former, because of all the usual buzzwords of "density" "power-efficiency" and "compute", and the latter, because notebooks are fundamentally obsolete from the moment they hit the market: Performance is strictly downhill from there, barring a miracle with MXM GPUs or adding in extra RAM. Maybe swapping spinning rust for flash, but even that is progressively less likely these days. They also need to get deals in with vendors for the server kit. Having super inexpensive 32C/64T chips that clock just as high as their Intel counterparts is worth next to nothing if you don't have HP/Dell/Cisco/IBM/etc. prepared to sell systems with them.
|
|
#
¿
Jun 4, 2017 08:02
|
|
|
SinineSiil posted:Are you sure? The worst (best?) part is you can't be totally sure these days.
|
|
#
¿
Jun 6, 2017 17:50
|
|
|
I wonder if there are some chips that really lost the silicon lottery and just don't have enough volts at stock clocks to keep up under a really intense long-running 100% load situation.
|
|
#
¿
Jun 9, 2017 07:42
|
|
|
wargames posted:Would zen2 with just a slight bump to volts all around be a lazy fix for such things or is zen 2 going to be on 7nm? This is entirely a wild-rear end guess based solely on my observations in both the CPU threads on SA, some threads on a few other boards, and conversations I've had about the CPU R&D process with some other folks in the industry, but I suspect the following will happen: Zen 1.0 (current Zen) will continue to receive stability and performance improvements through microcode built into AGESA. Zen 1.5 (some folks are calling it Zen+) will happen early next year; it will likely be a process refinement of the Zen architecture with all the microcode-based improvements intact, and then some. It'll be when AMD is comfortable with the process they're working with and designing for. I suspect a die shrink would be in the cards at some point if AMD can end friendship with GloFo, now Samsung is my best friend. There's no way they could do a competitive 10nm part on the GloFo process. Cannonlake (Intel's Kaby Lake die shrink) has been pushed back a bit last I checked.
|
|
#
¿
Jun 9, 2017 09:10
|
|
|
It's late, yeah, but in general AMD's platfrom and CPU documentation is usually phenomenal. There's a night and day difference between what's in the publicly-available Intel developer manuals and the publicly-available AMD developer manuals. AMD's BIOS and Kernel Development Guides are incredible. You can literally write firmware for their platforms with them, whereas if you want to do something even remotely similar with Intel you better be a board partner.
|
|
#
¿
Jun 14, 2017 23:34
|
|
|
I still think the problem is an ever so slightly (like, 10mV) too low stock voltage for the stock clocks for the chips that would usually be considered a loss in the silicon lottery.
|
|
#
¿
Jun 18, 2017 05:58
|
|
|
They ran the Intel systems with the RAM downclocked to 2133 MHz too? WTF?
|
|
#
¿
Jun 20, 2017 23:01
|
|
|
Subjunctive posted:Didn't AMD release a compiler that sucked? It was basically just LLVM with some patches.
|
|
#
¿
Jun 21, 2017 03:52
|
|
|
Paul MaudDib posted:What is NPT in this context? Nested page tables, another term for second-level address translation, or as Intel calls it, Extended Page Tables. It's a hypervisor acceleration technology that allows hypervisors to use the full feature set of x86 paging without the associated software overhead of page table shadowing. Hyper-V has required it since 2008R2, and bhyve and OpenBSD vmm both require it.
|
|
#
¿
Jun 23, 2017 03:12
|
|
|
So B350 looks good for a single GPU build. Any goon-preferred B350 mATX/ATX board? Or at least one that's known to be pretty good? I want to slam some 3600 MHz RAM into it and call myself good for the next 5+ years.
|
|
#
¿
Jun 30, 2017 21:13
|
|
|
FaustianQ posted:You know an interesting and kind of depressing thought, but maybe TR exists solely because AMD doesn't think they'll get enough orders for EPYC and are trying to recoup costs? I can't really explain my answer because of NDAs but this is definitely not the case.
|
|
#
¿
Jul 6, 2017 22:41
|
|
|
My dream is for them to get Zen on a process that does 4.5-4.8 GHz depending on lottery, and then do 8 core chips with 16 GB of HBM on die. Infinity Fabric at HBM grade speeds? Yes please. That is entirely a lofty dream though. My hopes are that Zen+ will be able to reliably hit 4.2 GHz with some extra volts and do 3200+ MHz DDR4 without a hitch. The difference in CPU performance on current Zen between 2400 MHz and 3200 MHz is absolutely astounding.
|
|
#
¿
Jul 7, 2017 07:45
|
|
|
incoherent posted:AMD Ryzen 7 1700X Processor - 299 on amazon prime day. It looks to be the lowest ever. Fuuuuuck. That's really good. It's too bad I just bought a new car 
|
|
#
¿
Jul 11, 2017 21:18
|
|
|
The really neat part of the CCX system is that what used to be "single core turbo" is now "single core per CCX turbo". So on a 16C/32T TR4 box, "single core turbo" is actually four cores active (one per CCX). At least that's my understanding of it. They really need to release the Zen BKDG already...
|
|
#
¿
Jul 14, 2017 07:11
|
|
|
Aww man I was hoping that Athlon would be a name for a higher end part. I want to feel like I'm back to the days of building a super fancy new Athlon based gaming PC. 
|
|
#
¿
Jul 14, 2017 20:14
|
|
|
That is a hell of a lot better than the "stock cooler" for Intel high end chips being both a price-gouged Asetek AIO and a completely separate purchase from the chip itself.
|
|
#
¿
Jul 18, 2017 06:04
|
|
|
I'm holding off on buying a 1700 now that we know that whatever Zen stepping Threadripper is on can in fact surpass 4.0 GHz. Hopefully it'll trickle down to the 1750 or whatever's up next.
|
|
#
¿
Aug 4, 2017 21:26
|
|
|
Well I just realized that AMD's Zen 2 sweet spot CPU will be called the 2600 
|
|
#
¿
Aug 4, 2017 21:55
|
|
|
KVM needs to fix their broken poo poo and AMD is too busy doing actually important things (like developing and releasing CPU architectures that don't suck for the first time in nearly a decade) to fix their broken poo poo for them. At least the KVM people aren't being openly hostile to AMD about it like Torvalds vs. Nvidia.
|
|
#
¿
Aug 9, 2017 09:04
|
|
|
Malcolm XML posted:Game developers discover ancient multicore secrets Engine developers: "We've invented this revolutionary new algorithm that'll improve multithreaded code on high core count systems!" Kernel developers: "Um, hey, about that..."
|
|
#
¿
Aug 9, 2017 20:19
|
|
|
I want to believe that sub-48CU Vega doesn't suck. I want to believe that they can shove 12 CUs into an APU, underclock them a bit, and make a 65W budget gaming monster out of it. Everything we've seen though is that Vega's a compute card that happens to have some rasterizers attached to it, and that more than anything else scares me about the future of Radeon. I love my GTX 1070 to bits -- getting 1080p144 in FFXIV on max settings is a thing of beauty -- but I would not want an actual no poo poo DOJ-intriguing Nvidia monopoly.
|
|
#
¿
Aug 9, 2017 21:49
|
|
|
Wirth1000 posted:Here's your first AIO that covers the entire dang thing for Threadripper. That is a gargantuan cold plate. I am incredibly interested in this. Unrelated, I'm thinking I need to re-paste my H60i. My i7-3820's temperatures are way too goddamn high (wtf is this 58 C idle).
|
|
#
¿
Aug 9, 2017 22:07
|
|
|
Despite being the OP of that thread I heartily encourage you to not get into it. It's a losing man's game and we're all just still posting in there to laugh at people who start getting into it now.
|
|
#
¿
Aug 10, 2017 01:42
|
|
|
 That seems a bit implausibly high. Like, "we forgot to subtract the GTX 1080" high, or "SB-E overclocked to 5 GHz at 1.5V" high.e: Actually something is fucky with that graph. It says the stock 7700K is pulling 156W with a GPU, and a stock 7700K is 91W? So that leaves 65W for platform + GPU. Even assuming the platform is only taking 15W, that leaves 50W for the GPU, and what even is that, a GT 730? Kazinsal fucked around with this message at 20:53 on Aug 10, 2017 |
|
#
¿
Aug 10, 2017 20:49
|
|
|
My stopgap until Zen2 is probably going to be hoping for a flood of super cheap second hand 7700Ks and Z270 boards. Gotta get off this i7-3820.
|
|
#
¿
Aug 11, 2017 00:21
|
|
|
PCIe 3.0 was finalized in 2010, boards started showing up around 2012 with Z77/Ivy Bridge. Their goal for finalizing PCIe 3.0 was 2008 IIRC so realistically 5.0 won't be until around 2023, and implementations around 2025. wargames posted:two differant teams, 4 had issues and 5 didn't. Why did we go from ipv4 to ipv6? ipv5 had issues. Plus this. You can actually read the "IPv5" spec though! It's RFC 1819, Internet Stream Protocol Version 2.
|
|
#
¿
Aug 30, 2017 06:12
|
|
|
VostokProgram posted:Aren't qemu and kvm related? I thought qemu by itself only does emulation, and you need kvm to use hardware assisted virtualization QEMU emulates hardware and has a dynamic translation CPU engine. KVM is a VT-x hypervisor that QEMU can frontend. QEMU can do emulation/dynamic translation of a shitload of platforms and architectures, while KVM is only able to be a hypervisor for the host platform (and iirc is only available on x86-64 and ARM64).
|
|
#
¿
Aug 31, 2017 21:14
|
|
|
You guys are going about this all wrong. AMD knows exactly how to push Intel's buttons now that Zen is actually competitive. Ladies and gentlemen... Zen Lake.
|
|
#
¿
Sep 6, 2017 06:56
|
|
|
KVM has known bugs that only manifest on AMD that they refuse to fix for some inane reason. This has been a thing for years, yet people still try to use KVM on AMD. Just install Xen.
|
|
#
¿
Sep 16, 2017 22:32
|
|
|
When you're GPU bottlenecked on Ryzen, RAM speed won't matter that much, but if you're CPU bottlenecked, then RAM speed can seriously shift the bottleneck back to the GPU. IIRC the numbers testing Deus Ex Mankind Divided on an R7 1700 with a GTX 1080 had LegitReviews (I think it was them) at 95 FPS with 2133 and 110 FPS at 3200 when testing at 1920x1080, but they were GPU bottlenecked at 2560x1440.
|
|
#
¿
Sep 17, 2017 00:55
|
|
|
CFL's gonna be my Christmas present to myself. No two ways about it.
|
|
#
¿
Oct 5, 2017 22:05
|
|
|
Hell of looking forward to Ryzen Espresso Edition that's just a single zen core boosting to 6 GHz. (someone would buy it I'm sure)
|
|
#
¿
Oct 6, 2017 20:41
|
|
|
AC:O's "anti-tamper" solution is to literally run the DRM in a non-hardware assisted, software-based x86 virtualization wrapper. Even if that's using some kind of magical dynamic translation engine that's still going to eat cores like a hot drat.
|
|
#
¿
Nov 1, 2017 18:09
|
|
|
Doing full software virtualization and throwing the hot path and friends in is absolutely mindblowingly retarded though. It's not quite "they just threw Denuvo in a QEMU instance" but it's drat close.
|
|
#
¿
Nov 1, 2017 18:20
|
|
|
|
| # ¿ Apr 19, 2024 21:18 |
|
|
SoftNum posted:For those still following the AMD NPT 10 year old bug, someone put together an actual patch (as opposed to a hack): Ten years of "AMD NEEDS TO FIX THEIR poo poo SCREEEEE" followed by one guy adding seven loving lines of code to fix the actual bug in the hypervisor. God drat.
|
|
#
¿
Nov 10, 2017 19:15
|
|