|
Palladium posted:Even 4 DIMMs now are a pretty worthless feature when you can have cheap 8GB sticks. http://arstechnica.com/gadgets/2015/08/intel-skylake-core-i7-6700k-reviewed/ It's also important to note that, while Haswell-E and X99 supported quad-channel operation, Skylake and Sunrise Point will only do dual-channel. Presumably Skylake-E will again bring back a quad-channel integrated memory controller. So basically all those quad channel DDR4 kits being sold for Skylake are of no benefit right now.
|
# ? Aug 17, 2015 09:23 |
|
|
# ? Apr 23, 2024 19:37 |
|
Cardboard Box A posted:So basically all those quad channel DDR4 kits being sold for Skylake are of no benefit right now. I thought that was pretty much common knowledge since forever? In other news, this is what happens when you get suckered by marketing: http://forums.anandtech.com/showthread.php?t=2443278 quote:Built a budget PC for the 1st time... quote:budget PC
|
# ? Aug 17, 2015 10:40 |
|
Oh boy, he already bought the board? That's even more hilarious He's gonna have trouble getting past 4.3ghz on the G3258 no matter what board he uses (which is why I suggest either a bare-featured H97 no more than $55 after discount or an H81 board), he might as well go full retard and get a 4790k and direct contact cool the CPU *crunch*
|
# ? Aug 17, 2015 10:42 |
|
Anime Schoolgirl posted:Oh boy, he already bought the board? That's even more hilarious There something about 4+ GHz that drives builders into retard OC mode. Anyone could MCE a 4690K @ 3.9GHz on all 4 cores on a H81 and *still* have spent the same $$$ as that guy.
|
# ? Aug 17, 2015 11:13 |
|
Boogaleeboo posted:I suppose it depends on what you mean by 'littered'. Even most of the itx boards I've seen have a couple [Generally an A and a C], and the most I've seen is 6 3.1 slots on the ASUS Deluxe board. The majority of them have them except for the extreme budget options. That's a pretty good way to push adoption. Twerk from Home posted:It's pretty amazing that a chip @ 3.3 base / 3.7 turbo can keep up with two 4.0 base / 4.2 and 4.4 turbo CPUs in games. That's a pretty serious clock deficit. Combat Pretzel fucked around with this message at 11:44 on Aug 17, 2015 |
# ? Aug 17, 2015 11:39 |
|
USB 3.1 Gen 1, USB 3.1 Gen 2, USB Type C, and how many should support USB PD?
|
# ? Aug 17, 2015 11:41 |
|
I was just wondering what the hell the generational stuff is about, those idiots renamed USB 3.0? Christ.
|
# ? Aug 17, 2015 11:43 |
|
LiquidRain posted:USB 3.1 Gen 1, USB 3.1 Gen 2, USB Type C, and how many should support USB PD? All these needless differentiation of standards are just plain awful obstacles to widespread adoption, not to mention plain old USB3.0 is more than good enough for most consumer applications.
|
# ? Aug 17, 2015 11:55 |
|
Also, USB 3.0 doesn't allow hackers to gain complete control over your memory by plugging in a dongle running a Python script.
|
# ? Aug 17, 2015 12:15 |
|
The new Macbook will I guess be the big driver for USB-C adoption. It's cool that they can unify adaptors for a bunch of different things but seriously, only one usb port on the laptop itself? That's just gunna be annoying.
|
# ? Aug 17, 2015 12:37 |
|
dpbjinc posted:Also, USB 3.0 doesn't allow hackers to gain complete control over your memory by plugging in a dongle running a Python script. ASLR + 64 bit address space should make that more hassle than it's worth, shouldn't it?
|
# ? Aug 17, 2015 13:36 |
e: ignore
|
|
# ? Aug 17, 2015 13:46 |
|
Mr Chips posted:ASLR + 64 bit address space should make that more hassle than it's worth, shouldn't it? No. ASLR protects against buffer overflow attacks; they are referring to the Thunderbolt DMA vulnerability.
|
# ? Aug 17, 2015 14:35 |
|
sout posted:The new Macbook will I guess be the big driver for USB-C adoption. I'm thinking they might add another port or two, especially if this means they get rid of the Macbook Air line (currently the Macbook is lighter than the Macbook Air)
|
# ? Aug 17, 2015 15:09 |
|
MrPablo posted:No. ASLR protects against buffer overflow attacks; they are referring to the Thunderbolt DMA vulnerability.
|
# ? Aug 17, 2015 15:24 |
|
Combat Pretzel posted:At least in the case of Windows, it randomizes the addresses of just about everything, i.e. the base addresses of executables and DLLs, as well as the heap and various process metadata. It randomizes virtual addresses, but I think the DMA attack operates on physical addresses. And it doesn't randomize kernel addresses, so DMA can target the kernel heap and similar if it has trouble going virt to phys. (ASLR mitigates against more than just buffer overflow attacks, of course.)
|
# ? Aug 17, 2015 15:43 |
|
Mr Chips posted:64 bit address space How many bits now?
|
# ? Aug 17, 2015 15:46 |
|
Subjunctive posted:It randomizes virtual addresses, but I think the DMA attack operates on physical addresses. And it doesn't randomize kernel addresses, so DMA can target the kernel heap and similar if it has trouble going virt to phys.
|
# ? Aug 17, 2015 15:48 |
|
DrDork posted:TechReport's review has the 5775C in the mix. It generally beats out the 6700k at anything gaming related by small margins, but is ~20% slower in the non-gaming CPU tasks (compression, mostly). That's at stock clocks, of course, and it doesn't look like Skylake is gonna do much in the way of overclocking. Oh wow, Broadwell is absolutely kicking rear end with that giant cache even with such a clock deficit. It not only helps to raise average FPS, it noticeably reduced spikes in frame times, which is an amazing result. So what the hell is Intel up to? Now we finally have the direction in which to go: a giant cache along with a 4/4.4GHz clock speed should be something worth lusting after. Hell, the fact Broadwell is used in a Z97 board is the icing on the cake. It looks like the 5775C can overclock a decent amount, at least north of 4GHz, and definitely uses less power than Devil's Canyon doing it. At this point, I want that more than Skylake.
|
# ? Aug 17, 2015 16:17 |
|
HalloKitty posted:So what the hell is Intel up to? Now we finally have the direction in which to go: a giant cache along with a 4/4.4GHz clock speed should be something worth lusting after. Hell, the fact Broadwell is used in a Z97 board is the icing on the cake. It looks like the 5775C can overclock a decent amount, at least north of 4GHz, and definitely uses less power than Devil's Canyon doing it. At this point, I want that more than Skylake. Otherwise, they'll give us basically the same chip forever and any actual improvements will likely be one-offs like the L4 cache on Broadwell.
|
# ? Aug 17, 2015 16:25 |
|
Combat Pretzel posted:If it can only address physical addresses, how does it make things easier? I thought with all that virtual addressing and paging, the representation of it in physical address space is quasi random? Do you commonly have 512TB of physical memory populated?
|
# ? Aug 17, 2015 16:29 |
|
Combat Pretzel posted:If it can only address physical addresses, how does it make things easier? I thought with all that virtual addressing and paging, the representation of it in physical address space is quasi random? The way virtual memory works is that separate processes each operate in their own isolated virtual address space. This means that it's not possible for process A to access the memory used by process B, because process A has no way to refer to the memory for process B. For memory that is shared between processes (for example, kernel memory), the hardware (MMU) prevents unauthorized access (by generating a page fault). The kernel (and Thunderbolt devices) have physical (not virtual) access to memory, which means they have direct, unfettered access to the address space of any process or the address space of the kernel. So, the point is that DMA attacks bypass processor and operating system security restrictions. A malicious Thunderbolt device has access to everything in memory. This means the device can steal your private key, steal your grandma's chili recipe, or install malware, and neither the operating system nor the processor can do anything about it.
|
# ? Aug 17, 2015 16:45 |
|
MrPablo posted:The kernel (and Thunderbolt devices) have physical (not virtual) access to memory, which means they have direct, unfettered access to the address space of any process or the address space of the kernel.
|
# ? Aug 17, 2015 16:47 |
|
Subjunctive posted:It randomizes virtual addresses, but I think the DMA attack operates on physical addresses. And it doesn't randomize kernel addresses, so DMA can target the kernel heap and similar if it has trouble going virt to phys. and DMA attacks are mitigated by IOMMUs which have been built into intel platforms for years
|
# ? Aug 17, 2015 16:57 |
|
Anime Schoolgirl posted:Who the gently caress thought that was a good idea for anything you can just plug in? DMA is definitely not new or useless when properly implemented.
|
# ? Aug 17, 2015 17:00 |
|
JawnV6 posted:Do you commonly have 512TB of physical memory populated? Malcolm XML posted:and DMA attacks are mitigated by IOMMUs
|
# ? Aug 17, 2015 17:49 |
|
Has it been 100% confirmed that Broadwell-E will still be on the old X99 platform?
|
# ? Aug 17, 2015 18:15 |
|
It hasn't been confirmed that Broadwell E will even exist
|
# ? Aug 17, 2015 18:15 |
|
Anime Schoolgirl posted:Who the gently caress thought that was a good idea for anything you can just plug in? Because hypothetically, for some narrow range of applications, virtual address translation might be too slow and bottleneck the connected device. In reality, this is not likely outside of very specific scenarios for which you would probably buy specialty hardware anyway.
|
# ? Aug 17, 2015 19:17 |
|
Thunderbolt devices have the DMA vulnerability, yes. But they have another problem: they're PCIe devices. PCIe devices can have Option ROMs. Option ROMs are executed at boot by firmware. In other words, if you plug in a malicious device and do a cold boot, you can now execute arbitrary code that has full access to everything before the OS even starts. If the UEFI code is lovely enough, you can even reprogram the system firmware so that your malware can never ever be removed without physically modifying the motherboard. There are two protections against this: disable the Thunderbolt ports entirely (through UEFI, if possible, or by physically destroying the port), or never let your computer out of your sight. However, your computer was shipped to you or to a store without those safeguards in place. A particularly powerful organization can easily intercept the shipment, attach the device, and boot the computer once, permanently tainting the device without anyone's knowledge. This is much easier and more stealthy than physically opening the device and soldering chips to the motherboard, and it can be done en masse. As far as I know, Apple's firmware in particular is still vulnerable to this.
|
# ? Aug 17, 2015 20:08 |
|
HalloKitty posted:So what the hell is Intel up to? Now we finally have the direction in which to go: a giant cache along with a 4/4.4GHz clock speed should be something worth lusting after. Hell, the fact Broadwell is used in a Z97 board is the icing on the cake. It looks like the 5775C can overclock a decent amount, at least north of 4GHz, and definitely uses less power than Devil's Canyon doing it. At this point, I want that more than Skylake. The 5775C looks interesting to me too but haha what is availability
|
# ? Aug 17, 2015 20:15 |
|
Firewire 2 everybody Sidesaddle Cavalry posted:The 5775C looks interesting to me too but haha what is availability
|
# ? Aug 17, 2015 20:15 |
|
Combat Pretzel posted:What does that have to do with anything? Unless you have 512TB of physical memory, the physical space will be smaller than virtual. ASLR bypass often includes sniffing around for a known value. Sniffing around for a known value takes less time in a smaller space.
|
# ? Aug 17, 2015 20:19 |
|
Combat Pretzel posted:If it can only address physical addresses, how does it make things easier? I thought with all that virtual addressing and paging, the representation of it in physical address space is quasi random? If my understanding of the NT kernel allocator is correct, it will tend to fill from a fixed offset upwards. You're therefore very likely to find a given kernel structure in the first few dozen megs of physical memory. I don't recall if the kernel heap starts at a fixed address, but if not then the address at which it starts has to be stored in a fixed address. Similarly, you're very likely to find process memory allocated in the lowest regions of physical memory first. Of course, if you can read arbitrary virtual memory, you can typically find the heap without much difficulty by inspecting the mapping table to find the standard library or equivalent. e: Malcolm XML posted:and DMA attacks are mitigated by IOMMUs I haven't read the exploit, but my understanding is that thunderbolt attack in question isn't mitigated by the IOMMU. I'll take a look over lunch, though, because now I'm curious!
|
# ? Aug 17, 2015 20:26 |
|
This Black Hat presentation from 2013 indicates that at least at that point IOMMUs weren't in place on Apple hardware. Section 8 of this paper on DMA malware describes some of the limitations of IOMMUs as a mitigation, and ways that attackers can rendering them ineffective. Cool stuff!
|
# ? Aug 17, 2015 20:39 |
|
If you use see some random Thunderbolt monitor lying around your parking lot, don't rush in and plug it into your computuer
|
# ? Aug 17, 2015 20:41 |
|
WhyteRyce posted:If you use see some random Thunderbolt monitor lying around your parking lot, don't rush in and plug it into your computuer Yeah, I think it's more likely to be cases where someone has physical access and can interpose their thing on an existing monitor, or swap in a trojan ethernet adapter (similar, somewhat, to the random-USB-key attack).
|
# ? Aug 17, 2015 20:48 |
|
sout posted:The new Macbook will I guess be the big driver for USB-C adoption.
|
# ? Aug 18, 2015 00:41 |
|
japtor posted:The MacBook is just a start, eventually I'd expect the big driver to be replacing micro USB (phones, tablets, all those random devices that use USB power, etc). Hopefully. Micro USB can go to hell. What's so bad about micro USB?
|
# ? Aug 18, 2015 00:50 |
|
|
# ? Apr 23, 2024 19:37 |
|
Toast Museum posted:What's so bad about micro USB? It's so tantalizingly close to being omni-directional. But it's not. So you end up trying to jam the damned thing in the wrong way because it almost fits, and end up damaging it. I mean, not that I have ever done that, because I'm a careful and exacting person. But I have a wife and, uh, yeah. Thankfully so far it's only damaged the cables, and not the phone(s).
|
# ? Aug 18, 2015 01:04 |