|
Holy poo poo, fixing that cache clock stuff makes a huge difference. JFC. Take GTA5, parts of the city, the framerate dropped to 50 and sometimes less, now it's 70 and above.
|
#
?
Feb 1, 2018 16:16
|
|
|
|
| # ? Apr 26, 2024 16:16 |
|
|
Good to know. I know overclocking it doesn't matter, but it's one of those things like memory speed where you need enough and then it doesn't matter.
|
|
#
?
Feb 1, 2018 16:39
|
|
|
XP was the last OS to allow running 16bit installers 
|
|
#
?
Feb 12, 2018 03:54
|
|
|
Is there something MS specifically calls out as being unsupported with installers? 16-bit compatibility exists through Win10 32-bit.
|
|
#
?
Feb 12, 2018 13:36
|
|
|
BangersInMyKnickers posted:Is there something MS specifically calls out as being unsupported with installers? 16-bit compatibility exists through Win10 32-bit. No, people just don't bother to install non-64 bit Vista/7/8/10.
|
|
#
?
Feb 12, 2018 16:08
|
|
|
Are there even generic 16 bit installers? That is from before MSI (or installshield/wisescript) was a thing. Wouldn't every 16 bit installer be a custom executable/script that installs the application? Should run fine on 32 bit windows. E: Just remembered PIF files, but those haven't been supported in a long time.
|
|
#
?
Feb 12, 2018 21:50
|
|
|
NihilismNow posted:Are there even generic 16 bit installers? That is from before MSI (or installshield/wisescript) was a thing. Wouldn't every 16 bit installer be a custom executable/script that installs the application? Should run fine on 32 bit windows. InstallShield was first released in 1990, coinciding with the release of Windows 3.0 more or less, and there were several competitors too. And when Microsoft officially endorsed InstallShield for Windows 95 usage, it was often still a 16 bit installer until well into the 90s. That said yeah they run fine on 32 bit Windows 10 just as on 32 bit XP.
|
|
#
?
Feb 12, 2018 22:08
|
|
|
NihilismNow posted:Are there even generic 16 bit installers? That is from before MSI (or installshield/wisescript) was a thing. Wouldn't every 16 bit installer be a custom executable/script that installs the application? Should run fine on 32 bit windows. Many old, old school 32bit installers still had a tiny 16bit portion that displayed a custom error window when you tried to install a 32bit app on Win 3.x That little portion will make them hang up on modern systems. Oh god I just remembered win32s, please make the pain go away 
|
|
#
?
Feb 12, 2018 23:29
|
|
|
sauer kraut posted:Many old, old school 32bit installers still had a tiny 16bit portion that displayed a custom error window when you tried to install a 32bit app on Win 3.x Forgotten about NT 3.51 already? 
|
|
#
?
Feb 14, 2018 11:40
|
|
|
NT 3.51 was awesome.
|
|
#
?
Feb 14, 2018 19:54
|
|
|
VROC keys in the wild. Intel is apparently really going through with this and is dumb.  https://www.anandtech.com/show/12435/the-intel-ssd-dc-p4510-ssd-review-part-1-virtual-raid-on-cpu-vroc-scalability
|
|
#
?
Feb 16, 2018 01:46
|
|
|
Intel should try selling premium Spectre mitigation keys next.
mystes fucked around with this message at 01:52 on Feb 16, 2018 |
|
#
?
Feb 16, 2018 01:49
|
|
|
Must.. segment.. the market..
|
|
#
?
Feb 16, 2018 02:54
|
|
|
I really want to see someone benchmark it vs. mdadm and Storage Spaces.
|
|
#
?
Feb 16, 2018 03:13
|
|
|
So they announced it as a grand feature back with Skylake-X, but then decided to make them hard to get? The gently caress.
|
|
#
?
Feb 16, 2018 04:00
|
|
|
Has anyone taken one of those keys apart? I'd laugh if it was just a different configurations of resistors.
|
|
#
?
Feb 16, 2018 05:30
|
|
|
I hadn't seen that Intel pcie switch card before, kind of surprised it is x8. The PM8533 is a 48 lane switch and it's using (8 x 4) + 8 = 40.. Strange.
|
|
#
?
Feb 16, 2018 05:42
|
|
|
I'm trying to think of why a layer below storage spaces direct is desirable. Like, I'd prefer application-awareness of fault state and recovery, which is what you get with storage spaces
|
|
#
?
Feb 16, 2018 13:44
|
|
|
So when are we going to find these keys on Aliexpress for $3
|
|
#
?
Feb 16, 2018 15:46
|
|
|
KKKLIP ART posted:So when are we going to find these keys on Aliexpress for $3 Maybe never. They're almost certainly a small EPROM chip with a license key in there; Intel might be dumb, but they're not stupid. The bigger question is availability full stop: they were originally pointed as retail-channel add-ons, but they were basically never released and now might only come as part of OEM servers. I still haven't seen a lot of evidence that they are notably superior to any other software-based RAID setup, though, so still wondering what the intended use case is for this thing.
|
|
#
?
Feb 16, 2018 16:05
|
|
|
I wonder how hard they would be to reverse engineer and duplicate - it seems like Intel's nearly asking for that if they're selling a platform that advertises locked capabilities but doesn't sell the key to unlock them.
|
|
#
?
Feb 16, 2018 17:07
|
|
|
First Ice Lake samples are showing up in databases: http://ranker.sisoftware.net/show_system.php?q=cea598ab9ea792a694b2d5e8cebc81a7cef3d5bd80a6dee3c5a0c5f8c8ee9da098&l=en Lends credence to the rumors that Intel may skip Cannon Lake completely for most markets. Now Intel just has to, you know, have a releasable 10nm process.
|
|
#
?
Feb 16, 2018 17:42
|
|
|
Eletriarnation posted:I wonder how hard they would be to reverse engineer and duplicate - it seems like Intel's nearly asking for that if they're selling a platform that advertises locked capabilities but doesn't sell the key to unlock them. You assume anyone gives enough of a poo poo about some special-snowflake Intel soft-RAID to bother doing so.
|
|
#
?
Feb 16, 2018 18:08
|
|
|
excuse me sir but booting from your storage array is a very important configuration used by many institutions such as:
|
|
#
?
Feb 16, 2018 18:12
|
|
|
Eletriarnation posted:I wonder how hard they would be to reverse engineer and duplicate - it seems like Intel's nearly asking for that if they're selling a platform that advertises locked capabilities but doesn't sell the key to unlock them. How to reverse engineer and duplicate: Step 1) install literally any other software RAID
|
|
#
?
Feb 16, 2018 18:51
|
|
|
I'm fully expecting to see Intel find some way to detect if Storage Spaces is running and gimp it on machines that don't have that key installed.
|
|
#
?
Feb 16, 2018 20:15
|
|
|
DrDork posted:You assume anyone gives enough of a poo poo about some special-snowflake Intel soft-RAID to bother doing so. I assumed that there is potentially a reason for the feature to exist at all since I don't yet actually know anything about its performance (do you, or are you just being reflexively negative?) and I agree that the likely customer base is already aware of the existence of free software RAID solutions. It in fact doesn't seem totally crazy to me that Intel could come up with some kind of crazy CPU-based special sauce to improve performance when the storage volumes involved are directly attached to the CPU via PCIe lanes, but I'll freely admit that I don't know much about the fine technical details there and was just pondering in public. fishmech posted:How to reverse engineer and duplicate: Shocking revelation, thanks as always for sharing your wisdom fishmech. 
Eletriarnation fucked around with this message at 21:32 on Feb 16, 2018 |
|
#
?
Feb 16, 2018 21:27
|
|
|
Eletriarnation posted:I assumed that there is potentially a reason for the feature to exist at all since I don't yet actually know anything about its performance (do you, or are you just being reflexively negative?) and I agree that the likely customer base is already aware of the existence of free software RAID solutions. It in fact doesn't seem totally crazy to me that Intel could come up with some kind of crazy CPU-based special sauce to improve performance when the storage volumes involved are directly attached to the CPU via PCIe lanes, but I'll freely admit that I don't know much about the fine technical details there and was just pondering in public. Nope, it is actually just a soft-RAID program built into the processor. There is no special-sauce hardware, it is an x86 program that runs on the cores. The only advantages of this are (1) you can boot from your storage array, unlike soft-RAID, and (2) it doesn't consume a slot, unlike hardware RAID cards. It's super niche.
|
|
#
?
Feb 16, 2018 21:41
|
|
|
Paul MaudDib posted:Nope, it is actually just a soft-RAID program built into the processor. There is no special-sauce hardware, it is an x86 program that runs on the cores. The only advantages of this are (1) you can boot from your storage array, unlike soft-RAID, and (2) it doesn't consume a slot, unlike hardware RAID cards. Well, also it supports NVMe, which as I understand it no hardware RAID cards support yet. Intel is now shipping storage 1TB-8TB 2.5" NVMe drives, and if you want to run them in some type of RAID you sure as hell can't connect them to your old trusty LSI MegaRAID.
|
|
#
?
Feb 16, 2018 21:46
|
|
|
I was mostly curious about the nature of the DRM anyway, but I'm wondering more about the assertion that there's no unique benefit to the feature because it's all in software and implicitly equivalent to existing software RAID. It's well known that modern processors derive greater efficiency for many operations from breaking down assembly language into micro-operations internally instead of running the assembly code. Honest question - is it not possible that Intel would be able to add in extra code paths in their processors for RAID calculations which can benefit from working on a lower level than 3rd party code would have access to? I know that for the bulk of the operations there's probably a public and fully optimized way of doing it (bitwise XOR is pretty standard stuff) but does that apply to the entirety of the feature? Eletriarnation fucked around with this message at 22:04 on Feb 16, 2018 |
|
#
?
Feb 16, 2018 21:59
|
|
|
Completely software RAID is displacing the hell out of hardware RAID anyways. A hyperconverged platform these days like a Nutanix box consists of multiple blades, each with their own storage hooked directly to their onboard SATA controllers, and all the redundancy is done alongside the replication in software. They boot off a little 64 GB flash module that plugs straight into a SATA port, the failure plan for which is "you call Nutanix and they send you a new one that you swap in and the blade will reconfigure itself from the backups on the other blades". Intel could have made some money on this if they had gotten into it 15 years earlier, but proper software RAID is no longer a buggy mess and is becoming more and more commonplace even in the enterprise space. Eletriarnation posted:Honest question - is it not possible that Intel would be able to add in extra code paths in their processors for RAID calculations which can benefit from working on a lower level than 3rd party code would have access to? It's possible, but they don't. As far as anyone can tell, the dongle basically tells the processor "present a virtual RAID controller to the PCI bus with features X, Y, and Z". Intel's OS driver interprets that as it needs to. I would be incredibly surprised if there were more going on in those dongles than that. Kazinsal fucked around with this message at 23:21 on Feb 16, 2018 |
|
#
?
Feb 16, 2018 23:17
|
|
|
Yeah, ZFS is fantastic, mdadm is great, and all soft-RAID solutions prefer to just be handed drives directly. I wish all processors could bifurcate PCIe down to x4 width so you could just use NVMe adapter sleds instead of needing the kind with a PLX switch built in. Forget Virtual RAID on CPU, I just want Virtual JBOD on CPU. 
|
|
#
?
Feb 16, 2018 23:23
|
|
|
Paul MaudDib posted:Yeah, ZFS is fantastic, mdadm is great, and all soft-RAID solutions prefer to just be handed drives directly. Get servers with U.2 bays.
|
|
#
?
Feb 16, 2018 23:31
|
|
|
I'm looking at completing a new system build, but I'm hesitant to purchase the Xeon E3-1270v6 I've been eyeing up. Would I be better off waiting to see if a Spectre/Meltdown-corrected refresh CPU comes out?
|
|
#
?
Feb 21, 2018 04:26
|
|
|
They'll come out eventually, but probably not in the near future. If you need a system within the next 6 months (minimum), you're basically stuck with software/firmware workarounds like everyone else.
|
|
#
?
Feb 21, 2018 04:34
|
|
|
PUBLIC TOILET posted:I'm looking at completing a new system build, but I'm hesitant to purchase the Xeon E3-1270v6 I've been eyeing up. Would I be better off waiting to see if a Spectre/Meltdown-corrected refresh CPU comes out? Spectre fix will take years. Meltdown has all ready been corrected in the OS by forcing cache purges during sys calls. There are optimizations in the later gen of chips to optimize this use case an minimize the performance impact. I wouldn't bother waiting, it will be a while.
|
|
#
?
Feb 21, 2018 12:24
|
|
|
Didn't Intel say they'll have a fix in silicon this year? Still that was very vague so I wouldn't wait for it unless you're building an IO-intensive server.
|
|
#
?
Feb 21, 2018 14:34
|
|
|
mobby_6kl posted:Didn't Intel say they'll have a fix in silicon this year? Still that was very vague so I wouldn't wait for it unless you're building an IO-intensive server. The current-running assumption is that their answer was intentionally vague enough so that it covers the most likely scenario: that they've just incorporated the firmware patches into silicon, not that they've managed to re-engineer substantial chunks of the processor to fix the root problem.
|
|
#
?
Feb 21, 2018 14:43
|
|
|
I don't know what any of these words mean. Is this something new with Meltdown? https://arxiv.org/pdf/1802.03802.pdf
|
|
#
?
Feb 21, 2018 19:09
|
|
|
|
| # ? Apr 26, 2024 16:16 |
|
|
It's basically a tool to automate the creation of malicious payloads that exploit meltdown or spectre. Metasploit for CPU vulnerability exploitation.
|
|
#
?
Feb 21, 2018 19:23
|
|