|
calcio posted:Is the DoD really buying high end Asus motherboards and not Dell, etc.? Also, DoD does have the ability to just disable mass storage drivers for USB completely on all their physical machines that have access to a classified network, it's just that they'd rather just kill the USB stack entirely than bother trying because the process to excise just the mass storage drivers and scripts would be too much effort for the lazy pension-seekers and paper-pushers want to deal with.
|
#
?
May 11, 2012 22:25
|
|
|
|
| # ? May 9, 2024 10:38 |
|
|
necrobobsledder posted:The reference designs will be picked up by HP and Dell and if there's none that don't have a PS/2 port, they'll scream at Intel to put it back on because neither HP nor Dell want to pay for the engineering effort to add them back onto the reference they just briefly look at and send to Foxconn to build. Intel doesn't really have anything to do with that. It's the motherboard manufacturer paying $2.84 for a SuperIO chip like the NCT6776F and expending some design time in routing signals that would enable PS/2 functionality. My Precision still shipped with 2 serial ports and a single parallel port, which was pretty cool. I think it's a T3500 though, which may be a few years old at this point.
|
|
#
?
May 11, 2012 22:48
|
|
|
Anandtech got a hold of one of MSI's boards with Thunderbolt: http://www.anandtech.com/show/5829/a-first-look-at-thunderbolt-on-windows-with-msis-z77agd80
|
|
#
?
May 12, 2012 10:05
|
|
|
japtor posted:Anandtech got a hold of one of MSI's boards with Thunderbolt: Hot-plugging is going to need to have some good BIOS support as well. The tricky part about PCIe hot-plugging is allocating enough resources ahead of time for hot-plugging, in terms of buses and memory for the various BARs (memory, pre-fetchable, I/O space, etc).  Basically, in a regular non-hotplugging system, the BIOS will take care of enumerating all the PCI devices, and programming their configuration space registers (BARs, etc). The issue is with a hotplugged device, you are adding onto this existing enumeration in the OS. With PCI Express, your PCIe ports are really PCI-PCI bridges to software. A PCH PCIe port is a PCI-PCI bridge that may or may not have devices behind it. PCH PCIe ports sit on some bus number (usually 2) at device 28, functions 0 through 7. Assume that the PCH PCIe ports are sitting on bus 2. If say, we had a GigE controller plugged into one of these PCIe ports, that GigE controller would appear on its "own" bus, say as Bus 3, Device 0, Function 0. On top of needing its own bus number (the subordinate bus numbers need to jive with the assignments from BIOS, as seen above), the BIOS will also assign that devices the resources it needs, mostly memory and I/O space allocation based on the BARs the device presents. This is easy, because the device is plugged in at boot, and the BIOS is like "sweet bro, enumerating" and everyone is happy. If this was, say, an ExpressCard though, it gets a little trickier. The BIOS can make a safe bet about padding one bus for that PCH PCIe port, because no matter what, it'll always need at least one bus. (Downstream device needs its own bus #). But, it has to take a guess at what kind of memory resources to throw at that port. Most ExpressCard devices tend to be pretty "small"; simple SATA controllers or GigE controllers which are not demanding when it comes to MMIO or IO space needs. For TB though, maybe its a docking station where there's a USB controller, network controller and SATA controller? Or maybe it's just a single SAS adapter linked up with a x4 link? Or GPU. I can see it being kind of tricky with Thunderbolt (anyone who has some kind of Thunderbolt setup going right now should post lspci -t), but I also have an idea of a lazy implementation of BIOS support for it in my head (I just finished implementing something similar as well). Thunderbolt would use a downstream port on the PCH, Ports 0-3 or 4-7 ganged together in a x4, and tie that to the upstream port of the Thunderbolt controller on the remote end. I think that the TB controller implement a PCIe switch internally, meaning they break out into 4 x1 ports, if you want. You'd have to pad another four bus numbers for those, plus resources to support whatever could be there. This could get tricky though, because you could connect literally anything to that TB controller. It could get interesting when you have even more PCIe switches on the downstream end of the Thunderbolt connection. I've seen a bladed telecom system with over a hundred PCIe busses, and it can handle plugging-in/removing blades without choking at all.  The OS-side support is fun too. Windows XP wasn't actually PCIe-aware, but still handled ExpressCard hot-plugging OK thanks to ACPI-mediation. In say an XP system, when you plug-in an ExpressCard, the PCIe port will issue a SMI to the effect of "Hey guys, I'm training a link! Something AMAZING is happening!  " This would trigger something like ACPI _STA, and then the ACPI-aware OS would start looking for something new on the bus. Windows Vista and later, that actually know about PCIe, can do something a little different where the Windows PnP manager will directly handle the interrupt generated from the Presence Detect bit in SLTSTAT (Slot Status) and go hunt down the new thing you just plugged in. I wonder Linux users have been faring with Thunderbolt so far. (pciehp and its friends exist to support PCIe Hotplugging, but people still have issues apparently).Summary: PCIe hotplug can be tricky and we're at the mercy of the BIOS developers and the driver vendors. I wish FOSS developers could get sponsored to do more! e: I've got a lovely cold and I think I forgot like, a paragraph up there, so there's probably something that logically doesn't make sense above, sorry. movax fucked around with this message at 23:51 on May 12, 2012 |
|
#
?
May 12, 2012 23:20
|
|
|
That is a good post movax, thank you!
|
|
#
?
May 12, 2012 23:31
|
|
|
I would like to point out that Thunderbolt does suffer from the same security problems as FireWire and ExpressCard, in that any computer that has the ports enabled and automatically loads drivers for new hardware can have its memory read and/or altered by a malicious device. It's not a major issue, since it only affects non-server systems that need full disk encryption, which are a vast minority of systems out there, but it is something to consider for organizations where that would be an issue (i.e. DoD and friends).
|
|
#
?
May 13, 2012 05:25
|
|
|
dpbjinc posted:I would like to point out that Thunderbolt does suffer from the same security problems as FireWire and ExpressCard, in that any computer that has the ports enabled and automatically loads drivers for new hardware can have its memory read and/or altered by a malicious device. It's not a major issue, since it only affects non-server systems that need full disk encryption, which are a vast minority of systems out there, but it is something to consider for organizations where that would be an issue (i.e. DoD and friends). Oh yeah, totally. Anything attached to TB is at Ring 0 and can go to town on your PC. I thought there were rumours going around regarding Thunderbolt getting built into chipsets in the future, but at least for now the controllers will probably only appear on premium SKUs, and probably not at all on mass-business desktops.
|
|
#
?
May 13, 2012 06:36
|
|
|
Phone posted:Eh, I think that the main points for Ivy Bridge were to get their 22nm process working, practical tri-gate transistors, and to lower the overall thermal profile. I think that Ivy Bridge is going to be the "look ma, it works!" in the grand scheme of things, but it's still going to be able to hold its own.  edit: Since my post, I'm now seriously considered abandoning the laptop-only model and going 2500K/HD 6850 + dual core Ivy Bridge ultrabook. That'll be just under $2000, which is about my upper limit for a decked out Ivy Bridge laptop. Ugh, decisions. Josh Lyman fucked around with this message at 15:25 on May 13, 2012 |
|
#
?
May 13, 2012 11:56
|
|
|
Josh Lyman posted:As a laptop-only user, Ivy Bridge is meant for me, not you. I've been on a Core2Duo/GeForce 8400 machine for 4 1/2 years now and Ivy Bridge can't come soon enough. Wait until the XPS 13 goes Ivy <3
|
|
#
?
May 13, 2012 17:47
|
|
|
dpbjinc posted:I would like to point out that Thunderbolt does suffer from the same security problems as FireWire and ExpressCard, in that any computer that has the ports enabled and automatically loads drivers for new hardware can have its memory read and/or altered by a malicious device. It's not a major issue, since it only affects non-server systems that need full disk encryption, which are a vast minority of systems out there, but it is something to consider for organizations where that would be an issue (i.e. DoD and friends). In other news if you open up a computer and plug in a PCIe card you also get DMA access. And if you solder a different bios flash chip in you can bypass a boot password. DMA and guaranteed bandwidth is the entire point of those protocols.
|
|
#
?
May 13, 2012 21:32
|
|
|
movax posted: , probably doesn't help reading on a phone.From what I could get it makes me wonder about some of the current devices when Windows gets actual support for it (like hot plugging and everything). Right now it seems like some things are being tested against Windows just seeing TB as PCIe.
|
|
#
?
May 13, 2012 21:50
|
|
|
movax posted:Apologies if this is a dumb question, but is the situation any better with UEFI?
|
|
#
?
May 13, 2012 21:59
|
|
|
hobbesmaster posted:In other news if you open up a computer and plug in a PCIe card you also get DMA access. And if you solder a different bios flash chip in you can bypass a boot password. The point is plugging a device into someone else's macbook which is just lying around somewhere is a lot easier than opening their desktop and stuffing a new card in/soldering a new bios. Especially since you can do this while the computer is running and maybe be able to get something out of it even if it's only plugged in for a few seconds. Imagine cruising around a library with a worm payload on a flash drive back when all windows machines would autorun. Even if there are some restrictions to plugging devices without user interaction it doesn't prevent lending/borrowing of malicious devices. You're far more likely to borrow a tbolt device than borrow someone else's pcie cards or randomly solder poo poo to your computer. Longinus00 fucked around with this message at 22:06 on May 13, 2012 |
|
#
?
May 13, 2012 22:01
|
|
|
Toast Museum posted:Apologies if this is a dumb question, but is the situation any better with UEFI? I don't believe it is with respect to the nuts and bolts Movax posted, but it does have one huge advantage for developers: No more legacy code except what they specifically add in for support. With the standard BIOS system, backwards compatibility was being maintained all the way back to the original IBM PC, and every POST was done in a compatibility mode that ran the CPU, no matter how modern, like an Intel 8088. We're talking 16-bit processing, 1 MB of address space, ISA bus compatibility - all the legacy poo poo necessary to run DOS version 1.0 (assuming there are appropriate drivers for the system's actual hardware). Since, practically speaking, anybody who needs to run DOS on new hardware can virtualize a BIOS-based machine now, EFI lets system developers skip all that stuff. I don't know the nuts-and-bolts of the type of enumeration implementation that Movax talked about, but I would imagine that the increased firmware storage space and addressing that UEFI allows could be pretty drat handy. More room for BIOS-level logic to be stored in and operate.
|
|
#
?
May 13, 2012 22:32
|
|
|
Longinus00 posted:The point is plugging a device into someone else's macbook which is just lying around somewhere is a lot easier than opening their desktop and stuffing a new card in/soldering a new bios. Especially since you can do this while the computer is running and maybe be able to get something out of it even if it's only plugged in for a few seconds. Imagine cruising around a library with a worm payload on a flash drive back when all windows machines would autorun. Firewire did this too. Also I'm pretty sure this could be done with PCMCIA and ExpressCard on laptops! Maybe you should ask yourself why you'd need to worry about your friend giving you a malicious thunderbolt device?
|
|
#
?
May 13, 2012 22:54
|
|
|
Install Gentoo posted:Maybe you should ask yourself why you'd need to worry about your friend giving you a malicious thunderbolt device? Wouldn't the main risk be infected storage devices?
|
|
#
?
May 13, 2012 23:37
|
|
|
Toast Museum posted:Wouldn't the main risk be infected storage devices? Well did you ever worry about that when using a Firewire connected drive? If you have an infected Thunderbolt storage device, it's not likely to be able to steal the info someone wants, and then send it back to them unless you return the device to the malicious person. It's not likely that a malicious program installed on someone's computer could reprogram a thunderbolt device to execute an attack on its own either, and wait for the device to go into another computer to do it.
|
|
#
?
May 14, 2012 00:07
|
|
|
dpbjinc posted:It's not a major issue, since it only affects non-server systems that need full disk encryption, which are a vast minority of systems out there, but it is something to consider for organizations where that would be an issue (i.e. DoD and friends).
|
|
#
?
May 14, 2012 01:25
|
|
|
Install Gentoo posted:Firewire did this too. Also I'm pretty sure this could be done with PCMCIA and ExpressCard on laptops! Just imagine the security vulnerabilities of my currently set up MBP with a thunderbolt to expresscard bridge and two firewire expresscards installed! (all this to run 3 machine vision cameras off a laptop - someone needs to release a thunderbolt to PCIe bridge yesterday)
|
|
#
?
May 14, 2012 01:34
|
|
|
Toast Museum posted:Apologies if this is a dumb question, but is the situation any better with UEFI? I think development is easier with a UEFI-based BIOS solution, certainly, but the whole thing about needing to pad PCI buses and memory resources is due to the PCI spec, so there's no getting around that. It's just a tricky technical solution that requires development and a good amount of testing. I think Linux may get the shaft somewhat because FOSS developers only have so much time, and device vendors will spend time supporting their $$$ users on OS X or Windows. Most EFI-based BIOSes sport a CSM or Compatibility Support Module to handle booting non-EFI aware OSes and to maintain compatibility with software that expects legacy BIOS. Pretty much every PC BIOS will have this, Part of this can be things like CMOS emulation; your modern BIOS can simply store setup variables to NVRAM on the same SPI flash chip where the BIOS lives. Pulling a CMOS reset jumper would clear CMOS, but I'd bet the BIOS also reads that flag (CMOS Checksum bad/battery reset) and then loads failsafe defaults. e: I'm dumb and a fellow goon corrected me, of course Apple would have some type of CSM to allow for booting of Windows on Macs. movax fucked around with this message at 15:00 on May 14, 2012 |
|
#
?
May 14, 2012 03:40
|
|
|
So does intel plan on releasing any >4 core DESKTOP cpu's soonish? I know they have this i7 3930x beast But I really would rather not pay 600 for it. I do quite a bit of VMware so more cores is something I look for. AMD I know has the 8 core bulldozer but I haven't heard anything good from those. I am currently running an X6, but as my studies push higher my cpu usage is hitting a bit higher than I like it to, >95% usage with my clusters active is not uncommon. I might just need to look into biting the bullet and getting some server hardware in my computer.
|
|
#
?
May 14, 2012 05:01
|
|
|
Corvettefisher posted:So does intel plan on releasing any >4 core DESKTOP cpu's soonish? I know they have this i7 3930x beast But I really would rather not pay 600 for it. I do quite a bit of VMware so more cores is something I look for. AMD I know has the 8 core bulldozer but I haven't heard anything good from those. I am currently running an X6, but as my studies push higher my cpu usage is hitting a bit higher than I like it to, >95% usage with my clusters active is not uncommon. I'm no expert, but it really sounds like you're looking for a dual-socket Xeon rather than current AMD stuff. Also, I'd bet that a higher-end core i7 is quicker than your Phenom II X6, regardless of the 2 core differential.
|
|
#
?
May 14, 2012 05:36
|
|
|
For anyone who's interested, I just put together a 3570K system and it plays Starcraft 2 at 1920x1080 on high with no problems. Really impressed with the HD 4000.
|
|
#
?
May 14, 2012 06:06
|
|
|
Corvettefisher posted:So does intel plan on releasing any >4 core DESKTOP cpu's soonish? I know they have this i7 3930x beast But I really would rather not pay 600 for it. I do quite a bit of VMware so more cores is something I look for. AMD I know has the 8 core bulldozer but I haven't heard anything good from those. I am currently running an X6, but as my studies push higher my cpu usage is hitting a bit higher than I like it to, >95% usage with my clusters active is not uncommon. Yea, even the previous gen i7-2600K is faster in every single way than an X6 1100BE, including multithreaded applications that can take advantage of the X6's two extra physical cores. You could get two quad core Xeons for the same price as one 6 core chip, and you would need a new motherboard anyhow since you currently have an AMD board. The other thing is the high end desktop "K" chips don't support all the virtualization extensions (I think they do VT-d but not VT-x), since intel wants you to buy Xeons or non-overclockable non-K chips to do virtualization with.
|
|
#
?
May 14, 2012 13:20
|
|
|
This is interesting. A japanese site removed the heat spreader, cleaned up the CPU, applied non-crap thermal paste, and reattached the heat spreader. End result was some substantial drops in temperature and much better overclockability. Looks like intel needs to reconsider the thermal paste approach and go back to whatever they were doing in sandy bridge.
|
|
#
?
May 14, 2012 13:32
|
|
|
Goon Matchmaker posted:This is interesting. A japanese site removed the heat spreader, cleaned up the CPU, applied non-crap thermal paste, and reattached the heat spreader. End result was some substantial drops in temperature and much better overclockability. Looks like intel needs to reconsider the thermal paste approach and go back to whatever they were doing in sandy bridge. I think its more along the lines of the QA of the TMI installation was the same problems as Apple had/have with MBP thermal paste.
|
|
#
?
May 14, 2012 14:02
|
|
|
All TIMs perform within a few ºC of each other, it's mainly down to quality application. The best TIM put on very poorly will perform as bad as or possibly worse than an adhesive heat conductive pad put on correctly. Once you get past the application error variable, the range between the fairly generic white gunk TIM included with Coolermaster Hyper 212+/Evo, the aging Arctic Silver 5 (probably a poor choice for this application as it's conductive?), and the higher dollar TIMs like Noctua's NT-H1, AS MX-4, IC Diamond are literally a few degrees and you're more likely to get temp variance from putting it on wrong than anything else. Other factors like conductivity, ease of application, curing time, long-term viability (I hate TIMs that dessicate in 2-3 years, really hosed me with Arctic Silver 5 back when it was hot poo poo). I like Noctua's stuff for a good all-arounder that performs in the top percentile of TIMs in testing, is easy enough to put on right, and is priced competitively for the size of the tube, personally. But everyone has their favorite. I'm guessing they're using an inexpensive bulk non-conductive thermal grease like the Coolermaster pack-in stuff, and some will be better applied than others internally... Which, if I'm right, would explain unusually variable thermal performance chip to chip better than lithography issues. Bet it super-voids the warranty to pop the heatspreader off and put on your own TIM correctly, though 
Agreed fucked around with this message at 14:37 on May 14, 2012 |
|
#
?
May 14, 2012 14:34
|
|
|
Agreed posted:Bet it super-voids the warranty to pop the heatspreader off and put on your own TIM correctly, though I think the warranty goes out the window the second you think about whipping out the razor blade.
|
|
#
?
May 14, 2012 14:41
|
|
|
Even so, I strongly doubt they're going to change the production method just because. People are already getting good overclocks out of them, it's just rarer chips that can exceed 4.5GHz safely. That was pretty much the case with Sandy Bridge, and while superior thermal conductivity between the silicon and the heatspreader would probably make bigger overclocks more commonly viable, nonetheless Intel thinks with their wallet first and it's a pretty low margin change. The enthusiast market isn't their bread and butter, just a nice niche to have in their pocket, and they completely lack for competition there.
|
|
#
?
May 14, 2012 14:58
|
|
|
Am I the only one who has had a much easier time remembering Sandy Bridge and Ivy Bridge compared to previous Intel microarchitecture names? It feels like it's way easier to remember a name that's composed of two common words rather than one composed of a single uncommon word. It looks like they're going back to the old naming convention after Ivy Bridge and it saddens me.quote:Conroe/Merom
|
|
#
?
May 15, 2012 03:20
|
|
|
Cicero posted:Am I the only one who has had a much easier time remembering Sandy Bridge and Ivy Bridge compared to previous Intel microarchitecture names? It feels like it's way easier to remember a name that's composed of two common words rather than one composed of a single uncommon word. It looks like they're going back to the old naming convention after Ivy Bridge and it saddens me. Next two tock-ticks follow the same pattern.
|
|
#
?
May 15, 2012 03:35
|
|
|
Cicero posted:Am I the only one who has had a much easier time remembering Sandy Bridge and Ivy Bridge compared to previous Intel microarchitecture names? It feels like it's way easier to remember a name that's composed of two common words rather than one composed of a single uncommon word. It looks like they're going back to the old naming convention after Ivy Bridge and it saddens me. You should see AMD's naming scheme. The Piledriver microarchitecture will go into a number of products. The desktop performance market (Volan platform) will combine Vishera CPUs with Socket AM3+ boards/AMD 9x0 northbridges. The desktop mainstream market (Virgo platform) will match Trinity, Weatherford, and Richland APUs (based on the Fusion system architecture) with socket FM2. The Notebook mainstream and performance markets (Comal platform) will combine Trinity, Weatherford, and Richland APUs in a mobile package. Then there will be three server market segments: * Web/microserver (1 CPU): Dehli CPUs on Socket AM3+, same as Volan. * Energy-efficient servers (1-2 CPUs): Seoul CPUs on Socket C32. * Enterprise/mainstream (2-4 CPUs): Abu Dhabi CPUs with Socket G34. Piledriver, Volan, Virgo, Comal, Vishera, Trinity, Weatherford, Richland, Dehli, Seoul, Abu Dhabi. And that's just the 2012 releases of a single microarchitecture. The 2011 releases all had different names. The 2013 releases will have different names. Factory Factory fucked around with this message at 11:44 on May 15, 2012 |
|
#
?
May 15, 2012 04:04
|
|
|
Factory Factory posted:madness  WHYYYYYY
|
|
#
?
May 15, 2012 08:14
|
|
|
Cicero posted:Am I the only one who has had a much easier time remembering Sandy Bridge and Ivy Bridge compared to previous Intel microarchitecture names? It feels like it's way easier to remember a name that's composed of two common words rather than one composed of a single uncommon word. It looks like they're going back to the old naming convention after Ivy Bridge and it saddens me. Haswell, Broadwell, Skylake, Skymont. I guess they're not technically just two words, but I'll end up remembering them that way.
|
|
#
?
May 15, 2012 09:53
|
|
|
I had no idea that Intel was launching Ivy Bridge Xeons so soon after they launched the Sandy Bridge ones. http://www.engadget.com/2012/05/15/intel-launches-new-ivy-bridge-xeons-targets-microservers/ Granted, it looks like the Sandy Bridge ones are still the high end models as none of those are more than 4 cores. Also, with this announcement, it looks like the Dell R320, R420, and R520 all launched today. I guess I'll be spending my morning playing on dell's site to get a feel for these new machines.
|
|
#
?
May 15, 2012 13:36
|
|
|
Yeah, the E3 is the quad-core die. They're basically i5s or i7s with ECC RAM support. Sandy Bridge E3s were released very soon after the desktop SKUs were, as well (though these Ivys are even sooner, I think). Ivy E5 Xeons (for multi-processor machines) will no doubt take longer.
|
|
#
?
May 15, 2012 13:42
|
|
|
The next MBP upgrade should be huge: (not the power graph of the above benchmark but shows the potential energy/heat savings) 
|
|
#
?
May 15, 2012 14:21
|
|
|
bull3964 posted:I had no idea that Intel was launching Ivy Bridge Xeons so soon after they launched the Sandy Bridge ones. It looks like they also launched a whole pile of Sandy Bridge Xeons (according to Wikipedia at least, and they're what's in those Dells), the E5-2400 series and and E5-4600 (quad socket chips!). All three of those new Dells use the E5-2400, not the E3 chips. So no idea if there's anything using the E3 chips yet.
|
|
#
?
May 15, 2012 16:28
|
|
|
FISHMANPET posted:It looks like they also launched a whole pile of Sandy Bridge Xeons (according to Wikipedia at least, and they're what's in those Dells), the E5-2400 series and and E5-4600 (quad socket chips!). All three of those new Dells use the E5-2400, not the E3 chips. So no idea if there's anything using the E3 chips yet. Yeah, you're right, intel mentions them in the presser. I'm betting the E3 will go into precision workstations and maybe the R220 whenever they launch it.
|
|
#
?
May 15, 2012 16:38
|
|
|
|
| # ? May 9, 2024 10:38 |
|
|
I believe Xeon part numbering is supposed to be: Intel Xeon Processor Ex - ABCCDE x - 3, 5 or 7 A - max numbers of CPU per node B - Socket Type CC - Processor SKU D - option suffix for low TDP, etc E - Version suffix, e.g. 'v2' for IVB Xeons
|
|
#
?
May 15, 2012 16:44
|
|