|
mayodreams posted:The world has moved to annual and rolling updates, so you need to get with the program or be left behind.
|
# ? Jan 22, 2016 00:35 |
|
|
# ? Oct 4, 2024 12:13 |
|
adorai posted:I don't think you understand how business works. I need to provide my end users with a more or less homogeneous computing environment. Every workstation needs to be on more or less the same OS with the same featureset and options. My end users are not all able to adapt to even small changes in their desktop environment. Further, I cannot afford the manpower to do a rolling upgrade and retrain the end users to every OS upgrade cycle. Sure I do. One of the main reasons I left the previous job I had was the same mentality you expressed mostly because they hired people who couldn't use a computer, and the tradition was to use the lowest common denominator as the basis for any user facing technology. I agree that an environment should be as homogeneous as possible, but that should not stop you from doing updates and upgrades in a controlled fashion, which you should already be doing with an equipment replacement plan and cycle. Microsoft has been very aggressive to get everyone to Windows 10 to have a singular point of focus and support because the wide range of OS'es in the wild and still in support is a lot to manage. Which is also why they have 3 levels of update paths for Windows 10:
For reference: https://technet.microsoft.com/en-us/library/mt598226(v=vs.85).aspx Just because your company works a certain way doesn't mean everyone else does. Most of the places I've worked at have had solid upgrade plan with regard to new OS and hardware cycles.
|
# ? Jan 22, 2016 04:09 |
|
mayodreams posted:Just because your company works a certain way doesn't mean everyone else does. Most of the places I've worked at have had solid upgrade plan with regard to new OS and hardware cycles. Sure, maybe the jobs you've worked for. But the majority of large corps do NOT migrate OSes in any timely manner for any numbers of reasons.
|
# ? Jan 22, 2016 13:43 |
|
Even the majority of small corps don't, either. I still have one legacy app that won't run properly on anything newer than 7, and only with the classic shell. Wouldn't mind ditching it, except that the work done in that program fully represents 95% of all the CAD we've ever done.
|
# ? Jan 22, 2016 14:40 |
|
Stanley Pain posted:Sure, maybe the jobs you've worked for. But the majority of large corps do NOT migrate OSes in any timely manner for any numbers of reasons. And how many companies that don't migrate OSes in a timely manner are really diligent about buying brand new CPUs constantly, again?
|
# ? Jan 22, 2016 16:13 |
|
fishmech posted:And how many companies that don't migrate OSes in a timely manner are really diligent about buying brand new CPUs constantly, again? My last place of employment estimated that their think pads lasted an average of 3 years. It's sort of alright.
|
# ? Jan 22, 2016 17:03 |
|
fishmech posted:And how many companies that don't migrate OSes in a timely manner are really diligent about buying brand new CPUs constantly, again? 3 year HW cycle tends to be the norm. But we're not talking about HW refreshes, we're specifically talking about how "every company you've worked for migrates their OS on a regular basis" Most corps have "recently" moved to Windows 7 and if you don't think that's going to be the norm for at least another 7+ years....
|
# ? Jan 22, 2016 19:12 |
|
Boiled Water posted:My last place of employment estimated that their think pads lasted an average of 3 years. It's sort of alright. ...somebody pulled an impressively fast one on the beancounters. Not sure if IT guys or IBM/Lenovo's corporate sales. Maybe both?
|
# ? Jan 22, 2016 19:16 |
|
SwissArmyDruid posted:...somebody pulled an impressively fast one on the beancounters. Not sure if IT guys or IBM/Lenovo's corporate sales. Maybe both? I'm not quite sure but it's probably a little column A, little column B. Breaking computers fast enough to replace them every three years is also super easy when they're all laptops and science personnel and sales people don't give a poo poo.
|
# ? Jan 22, 2016 20:04 |
|
yeah laptops get end-user'd at a frightening rate. desktops, not so much.
|
# ? Jan 22, 2016 20:14 |
|
SwissArmyDruid posted:Even the majority of small corps don't, either. Eh, some of this falls on the business, some on the vendor. If a company is still using an outdated, unsupported legacy application, they need to re-evaluate and start looking for something future-proof. Then again, that requires spending money, and lots of people running these companies don't ever like hearing about costs, regardless of how their business is performing. I understand not upgrading for compatibility or stability concerns - case in point with Windows 10, we've told our clients to avoid the upgrade until it's been tested across all systems and we know it'll work. But when it comes down to a money issue, or people being stubborn, incompetent, or just plain lazy, that's some grade-A choice bullshit in my mind. Example, I've dealt with past clients at my job telling us they don't want any PCs patched for security, then call us 5 times a month when one (or more) of their PCs gets an infection that used a security hole that would've been blocked if they'd updated. BOOTY-ADE fucked around with this message at 22:21 on Jan 22, 2016 |
# ? Jan 22, 2016 20:34 |
|
SwissArmyDruid posted:...somebody pulled an impressively fast one on the beancounters. Not sure if IT guys or IBM/Lenovo's corporate sales. Maybe both? I go at least 4 years in between laptop refreshes
|
# ? Jan 22, 2016 20:39 |
|
I've found three years is a good guide if A. you're giving these to non-IT folk and B. the laptops have regular hard drives and not SSDs. There's a small pile of 2012 era laptops at my work that'd do great for general office work with an SSD installed but they're a little banged up and handing one to a new employee gives a bad impression so new laptops it is.
|
# ? Jan 22, 2016 21:46 |
|
WhyteRyce posted:I go at least 4 years in between laptop refreshes Safe bet you know how to take care of your equipment though - most end users don't give a poo poo because it's not their own computer and they assume the office has extras sitting around. One of my past help desk jobs was great for that, idiots dropping laptops or spilling stuff on them, they they'd get pissed when we pulled a 2 year old beater from the closet while we repaired their machine. Sorry dude, you hosed up, you get the same treatment everyone else does - especially when you don't admit that you hosed up or had an accident.
|
# ? Jan 22, 2016 22:25 |
|
I'm using a laptop that was released 6 years ago. It runs fine, including stuff like TF2. Obviously on really heavy-duty compute tasks my 4690k outperforms it, but it's not something I worry about. There hasn't been a killer app for vastly more powerful computers recently (particularly laptops), and the performance increases have been fairly incremental. Mostly the improvements have been in efficiency and better low-power states. And SSDs have made a huge difference in perceived performance during that timeframe too. Even garbage hardware runs OK with a SSD. Heck, Windows 10 is measurably faster than Windows 8.1 and it stomps on 7. Hybrid boot is a massive improvement, I am at the desktop before I can reach up and turn my monitor on. I don't use fast boot on my laptop but it still boots up in about 5-10 seconds once it POSTs, even with an i7-720qm. Paul MaudDib fucked around with this message at 22:39 on Jan 22, 2016 |
# ? Jan 22, 2016 22:31 |
|
Ozz81 posted:Eh, some of this falls on the business, some on the vendor. If a company is still using an outdated, unsupported legacy application, they need to re-evaluate and start looking for something future-proof. Then again, that requires spending money, and lots of people running these companies don't ever like hearing about costs, regardless of how their business is performing. AHAHAHAHAHAHA- Oh, I totally get it. I've been fighting for us to move OFF that poo poo for ten goddamn years. It is only in the past year that we've started doing our CAD in Solidworks and CAM in HSMXpress, instead of just using Solidworks as a glorified document reader. (My boss doesn't make any sense 90% of the time, I know.) The only reason why I'm even able to do just that is because I blew him off and just started working in the free version of HSMXpress until I can convince him of its worth so that we need the non-free version. ( I can hear him say that the free version is good enough, why would we need the 3D milling features.)
|
# ? Jan 22, 2016 22:41 |
|
Ozz81 posted:Safe bet you know how to take care of your equipment though - most end users don't give a poo poo because it's not their own computer and they assume the office has extras sitting around. One of my past help desk jobs was great for that, idiots dropping laptops or spilling stuff on them, they they'd get pissed when we pulled a 2 year old beater from the closet while we repaired their machine. Sorry dude, you hosed up, you get the same treatment everyone else does - especially when you don't admit that you hosed up or had an accident. No lie, in a past job I considered swinging for Toughbooks for everyone, but that plan got shot down because loving Intel and their ultrabook craze was just starting to take off. About $12,000 in dead ultrabooks later....
|
# ? Jan 22, 2016 22:44 |
|
SwissArmyDruid posted:No lie, in a past job I considered swinging for Toughbooks for everyone, but that plan got shot down because loving Intel and their ultrabook craze was just starting to take off. I'm using the first work IT approved Ultrabook right now. Some Toshiba Ivybridge piece of garbage. The space bar doesn't really work too well (one side of it doesn't press), the screen is flexy and bendy, the whole thing is cheaper than normal feeling plastic. I really am amazed it's still going. Apparently there aren't any more of them left in supply so if I break it I can probably get a new one, but that would hurt my nerd pride Paul MaudDib posted:I'm using a laptop that was released 6 years ago. It runs fine, including stuff like TF2. Obviously on really heavy-duty compute tasks my 4690k outperforms it, but it's not something I worry about. People say this but my laptop always feels like poo poo at the end of it's refresh cycle even though my last two have had SSDs. I'd blame that more on the IT build and cornucopia of McAfee apps than the hardware though.
|
# ? Jan 22, 2016 23:00 |
|
SwissArmyDruid posted:
Lemme venture a guess: Your entire business model is based on being able to mill 3D things?
|
# ? Jan 22, 2016 23:38 |
|
WhyteRyce posted:I'm using the first work IT approved Ultrabook right now. Some Toshiba Ivybridge piece of garbage. The space bar doesn't really work too well (one side of it doesn't press), the screen is flexy and bendy, the whole thing is cheaper than normal feeling plastic. I really am amazed it's still going. Apparently there aren't any more of them left in supply so if I break it I can probably get a new one, but that would hurt my nerd pride Ah yes, MTBF. The REAL measure of an IT specialist's epeen.
|
# ? Jan 22, 2016 23:40 |
|
Stanley Pain posted:Sure, maybe the jobs you've worked for. But the majority of large corps do NOT migrate OSes in any timely manner for any numbers of reasons. Correct. What my point is that companies that do not adapt to modern technology are at risk for security and productivity issues. Like others have stated, there are no shortage of poorly managed companies that will feel the pain of this migration to faster paced releases because they are slow to adapt, lazy, incompetent, or too cheap. None of those reasons should stop MIcrosoft or Apple from releasing updates to their operating systems because some companies won't play ball. From my own experience, the previous employer I mentioned was running XP on 98% of user machines, and the vast majority of them were running a mix of Office 2003 and 2007. The reason they were not upgraded to be consistent was lovely management and leadership and the promise of 'we'll upgrade next year' and next year never came. That is how you end up running a half billion dollar business on technology that basically was not updated past about 2005 in YOOL 2015. The cost of migrating email, applications, desktops/laptops, servers, storage, and directory systems in both dollars, effort, and personnel was extremely high for this company. I, as well as a number of my colleagues, quit as a direct result of of being burned out after a year and a half of constant system failures and migrations that should have really taken 2-3 years to accomplish, but because everything was in such dire straights, it was non-stop. So when you have a revolving door of senior level engineers and admins, having any kind of continuity, or even getting talented people in after garnering a reputation can be extremely difficult, and can cripple an organization. These old and out of touch managers and execs also tend to hate subscriptions, which is the way the world is moving for software, like it or not. Microsoft makes no qualms about calling Windows 10 'Windows as a Service.' They only support the last 2 versions of Outlook/Office for Office365 functionality, which was an aggressive way to root out the assholes using XP because Office 2010 was the last supported version of Office on XP, and with the release of 2016 last year, it would no longer be supported. Not to mention it barely worked most of the time anyway because XP stopped getting updates that the other versions were getting that resolved issues with authentication and licensing. There will always be resistance to change as it is human nature. But when you have to consider security, productivity, and cost associated with compliance, the days of buying Server 2003 or XP and running it past the supported dates are over if you want to ensure your technology has the best chance of avoiding being compromised. IT is no longer a cost center, and should be considered a critical part of the success of the business, and if it is not, you're gonna have a bad time.
|
# ? Jan 22, 2016 23:52 |
|
How would you recover a company like that if they were hit with something nasty, like say, crypto locker across all networked xp-shitboxes?
|
# ? Jan 23, 2016 00:40 |
|
mayodreams posted:IT is no longer a cost center, and should be considered a critical part of the success of the business, and if it is not, you're gonna have a bad time. Boiled Water posted:How would you recover a company like that if they were hit with something nasty, like say, crypto locker across all networked xp-shitboxes?
|
# ? Jan 23, 2016 00:57 |
|
adorai posted:That doesn't mean you have to piss away money. lol. there were no backups.
|
# ? Jan 23, 2016 01:29 |
|
Nobody cares about your awful company and the morons it hires to admin its it policies In other news what do people think of Polaris vs pascal?
|
# ? Jan 23, 2016 01:48 |
|
WhyteRyce posted:People say this but my laptop always feels like poo poo at the end of it's refresh cycle even though my last two have had SSDs. I'd blame that more on the IT build and cornucopia of McAfee apps than the hardware though. For sure. My GF and I have identical laptops. Kaspersky gets an A+++ three thumbs up would-purchase-again from me. Super lightweight and super aggressive, and we paid a total of $20 for a 5-device license. We both run uBlock Origin and Malwarebytes too. McAfee and Norton are both 100% garbage that I refuse to allow on my systems. ESET is OK too, though. We do have (ancient) discrete GPUs, which really really helps. We have Quadro FX 880Ms, which is a 48 core 2GB Tesla card. Totally garbage by modern standards but it has good driver/API support and a lot of dedicated VRAM. I originally got mine for doing CUDA dev work and I bought her one when I found out that it actually was very decent for playing games too. Like I said, it's very playable on games like TF2 even at full resolution. If you keep your settings low it still does pretty OK even today. I actually have 16GB of memory and a slightly slower processor, she has 8 GB and a slightly faster processor. In practice the mobo supports up to 32GB. We both boot off 240 GB SSDs, but I run a secondary HDD caddy in my bay, while she runs the CD/DVD drive. We have USB 3.0 which is really nice for overall system performance feel, plus a fast-charge USB 2.0. We also have 900p screens, a magnesium chassis, and an absolutely rockin' keyboard. Also we still have at least one expansion option open - the ExpressCard slot. Thinkpads own, even refurb. I think I paid $250 for hers about a year ago, and then tossed another $100 or so in upgrades into it. We're both absolute hell on laptops and so far they both have done extremely well. They're not fast by modern desktop standards but they don't feel like the 6-year-old junkers they are and they wear like hell, which is great for a laptop. I need to unlearn the habit of picking it up by the screen because I know the next laptop won't be OK with that. I've dropped it repeatedly and the most I've done is knock out a single link of the plastic exhaust grill that protects the cooling module. She likes to let the processor nuke itself inside the bed covers. Paul MaudDib fucked around with this message at 04:35 on Jan 23, 2016 |
# ? Jan 23, 2016 04:08 |
|
Malcolm XML posted:In other news what do people think of Polaris vs pascal? See my comments in the GPU thread and an interesting response from Durinia. tl;dr: the NVIDIA presentation at the CES show was awesome and super interesting but it also released hard numbers for the Drive PX2 which lets you drill down on the performance of GP106. I think it's shaping up to be a 50% boost in performance, wild-rear end guess. However, AMD has already done an HBM-based arch and has a significant advantage in that they're partnering with Samsung, who is basically just behind Intel in tech and is capable of going from raw materials to finished packaged chips. TSMC is just another chip maker. In theory the interposer concept unlocks a fuckload of derivative techs - multi-die chips, heterogeneous chips, etc, all tied into HBM. If AMD/Samsung can exploit that it could be a very promising platform in 2 years or so. Intel should be suing the pants off AMD right now because the GloFo/Samsung partnership is exactly the kind of thing that the (highly exploitative) x86/AMD64 transfer agreement forbids, because it poses an actual threat to Intel's market dominance in a variety of fields. Not just GPUs and CPUs, but eventually Greenland APUs could compete with Knight's Landing and stuff. Paul MaudDib fucked around with this message at 04:49 on Jan 23, 2016 |
# ? Jan 23, 2016 04:42 |
|
Malcolm XML posted:Nobody cares about your awful company and the morons it hires to admin its it policies I've said in the past that we were never going to see exactly what GCN was ever capable of until it got the die shrink it needed several years ago, and that as long as AMD insisted on using the scale tool and lopping off just enough bits to make their dies still be a reasonable size at 28nm, they were going to forever be fighting with one arm tied behind their back. I look forward to seeing if I was correct. I also remember the "wood screw incident". I have signed a six month contract with an artisanal sea salt maker that lives just down the coast in Monterey. I figure that's about as much salt as I'll need to take with anything that Nvidia says. I look forward to the actual parts coming out, and seeing people start taking them apart with a scalpel. Also, I want to see if NVidia makes any mistakes that AMD avoided by virtue of having had one generation's worth of experience with HBM.
|
# ? Jan 23, 2016 06:54 |
|
Paul MaudDib posted:Intel should be suing the pants off AMD right now because the GloFo/Samsung partnership is exactly the kind of thing that the (highly exploitative) x86/AMD64 transfer agreement forbids, because it poses an actual threat to Intel's market dominance in a variety of fields. Not just GPUs and CPUs, but eventually Greenland APUs could compete with Knight's Landing and stuff. I'm currently speculating a rather exploitative relationship between Samsung and AMD right now, as Samsung is super interested in getting away from Mali last I heard and starting up their own GPU division. AMD might be getting access to the second best fabs with everything it needs, but Samsung might be getting all the institutional knowledge it needs regarding GPUs, especially GPU designs utilizing HBM (so SoCs that strictly use HBM for graphics and system memory). I don't know what benefit Samsung gets from the CPU division precisely, that really depends on what elements K12 itself is leveraging. I'm sure AMD doesn't care, the advantages in the short term are way too high compared to the long term advantages for Samsung. AMDs focus is putting a new foundation on the house, can't really be quibbling over the mortgage right now.
|
# ? Jan 23, 2016 09:06 |
|
Paul MaudDib posted:See my comments in the GPU thread and an interesting response from Durinia. Question: is heterogeneous chips meant to refer to CPUs with GPUs onboard, or CPUs with related but lower power CPUs on board? Because at least for the latter, results in phones have not been promising.
|
# ? Jan 23, 2016 15:18 |
|
fishmech posted:Question: is heterogeneous chips meant to refer to CPUs with GPUs onboard, or CPUs with related but lower power CPUs on board? Because at least for the latter, results in phones have not been promising. Heterogeneous chips as in HSA? I've only ever seen it referred when talking about a CPU+GPU combination since CPUs and GPU operate very differently and thus it's actually heterogeneous, otherwise you might as well refer to the C2Qs as the first heterogeneous chips. EmpyreanFlux fucked around with this message at 19:03 on Jan 23, 2016 |
# ? Jan 23, 2016 17:00 |
|
Paul MaudDib posted:Thinkpads own, even refurb. I think I paid $250 for hers about a year ago, and then tossed another $100 or so in upgrades into it. We're both absolute hell on laptops and so far they both have done extremely well. They're not fast by modern desktop standards but they don't feel like the 6-year-old junkers they are and they wear like hell, which is great for a laptop. I lucked out at my last job, they'd get rid of computers for recycling every couple years and usually the only problems they had were the warranty being expired. Our bosses were cool and would let us go through the old equipment and nab working stuff if we wanted, so I nabbed a T61 one year, then later nabbed a T410 that I use daily. Both are great little machines and have taken a beating, probably the most durable I've owned so far. Only thing I did to both was add hard drives and beef up the RAM, they both run like clockwork
|
# ? Jan 23, 2016 17:44 |
|
FaustianQ posted:Heterogeneous chips as in HSA? I've only ever seen it refereed when talking about a CPU+GPU combination since CPUs and GPU operate very differently and thus it's actually heterogeneous, otherwise you might as well refer to the C2Qs as the first heterogeneous chips. C2qs had the same processor cores....
|
# ? Jan 23, 2016 18:04 |
|
Yea in the context of the PC market I would always assume the phrase "heterogeneous chip" to refer to some sort of CPU+GPU on die/package set up. Interposer chat: any reason why the interposer should have to be entirely under the GPU or CPU die? I'm no EE or GPU architect but most of the labeled die shots I've seen for modern GPU's like Hawaii or Fiji have the memory interface along the edges of the GPU. The positioning for those particular GPU's isn't ideal for this idea because the memory interface is along 3 of the 4 edges of the die...but what if they could move all of the memory interface to 1 side of the GPU die? Couldn't they then just make a interposer for those areas to the memory and then use "regular" packaging methods to deliver power and other I/O for the chip? I would think 1 small interposers would be cheaper than 1 gently caress off huge one. As Durinia noted in the post Paul MaudDib linked an interposer is really just a big chip in itself, just on a older and larger geometry "cheap" process, so silicon costs and yield savings typical for any other chip would apply and I would think to be significant if not substantial. The rub here is of course that using a small interposers + "normal" PBGA packaging might drive the defect rate through the roof for all I know. Or be totally impossible for whatever reason. Anyways if my description is off here is a mspaint of the idea: alternatively: goin' for actual bankrupt broke
|
# ? Jan 23, 2016 18:08 |
|
PerrineClostermann posted:C2qs had the same processor cores.... C2Qs were two separate dual cores dies, my point more or less that BIG.little and Tegra fall into the same vein of homogeneous computing within the realm of Fishmechs original question (CPU + CPU). SoCs in general are capable of heterogeneous computing as far as I am aware, but this is on CPU+GPU grounds. Xeon Phi strikes somewhere between a CPU and GPU as it's basically tens of little x86 cores working in parallel, IIRC. That's not HSA either, but it's direct competitor. PC LOAD LETTER posted:Yea in the context of the PC market I would always assume the phrase "heterogeneous chip" to refer to some sort of CPU+GPU on die/package set up. Maybe it'd down to fabrication and application simplicity equaling or trumping cost? Your first MS pain demo shows a good idea for APUs at least is likely what an HBM enabled APU will look like. EmpyreanFlux fucked around with this message at 19:56 on Jan 23, 2016 |
# ? Jan 23, 2016 19:19 |
|
Stanley Pain posted:Most corps have "recently" moved to Windows 7 and if you don't think that's going to be the norm for at least another 7+ years.... Basically, again, since Windows 7 happened, no Windows has provided an actual benefit to the end user that matters. Before you complain, under-the-hood performance increases and security improvements do not matter to users. As an end user it would have been convenient if 7 had become the last Windows ever made and were now in eternal life support mode.
|
# ? Jan 23, 2016 19:44 |
|
Basically 10 will become standard when all the win 7/8/8.1 shitboxes (which replace win XP shitboxes) finally die in 2025.
|
# ? Jan 23, 2016 19:46 |
|
blowfish posted:Basically 10 will become standard when all the win 7/8/8.1 shitboxes (which replace win XP shitboxes) finally die in 2025.
|
# ? Jan 23, 2016 20:37 |
|
fishmech posted:Question: is heterogeneous chips meant to refer to CPUs with GPUs onboard, or CPUs with related but lower power CPUs on board? Because at least for the latter, results in phones have not been promising. CPUs with heavy-duty GPUs onboard, i.e. the arch AMD outlined for HPC with a combination of Zen and Greenland. And also stuff like FPGA modules on a chip, which is Intel's new gimmick. Obviously there are already APUs out there and Intel is supposed to be shipping FPGA APUs sometime early this year. But having specialized dies allows you to have chips with more transistors and more optimal silicon processes, while also tying everything into fast on-package memory. Intel is also working on Knight's Landing, of course, which looks like a hell of a chip. That's the new Xeon Phi architecture. Basically it's a bunch of Atom cores on a package with low-latency memory dies, except the Atom cores have AVX512 so they can switch into a GPU-like SIMD mode or back to scalar processing anytime they want. So basically like a GPU that you built out of a CPU. It's not quite as fast in SIMD mode as a real GPU processor, but it could be much easier to integrate into standard processing. re CPU + CPU, i.e. big.LITTLE: It's a promising combination on paper, but my understanding is that it's hard to get a chip that does a good job of balancing real-world loads. It takes a lot of energy to switch between the cores, and it's easy to get "thrashing" back and forth between the processor types. Paul MaudDib fucked around with this message at 21:30 on Jan 23, 2016 |
# ? Jan 23, 2016 21:01 |
|
|
# ? Oct 4, 2024 12:13 |
|
PC LOAD LETTER posted:Yea in the context of the PC market I would always assume the phrase "heterogeneous chip" to refer to some sort of CPU+GPU on die/package set up. Yeah, if you were prepared to deal with the design consequences you should be able to scale out a processor arbitrarily like this. Obviously the interposer isn't free and you're talking about a lot of individual chips and assembly. Assembly costs may overtake the savings from yields at some point. Also you're talking about increasing the latency of various parts of the chip talking to each other, your overall clock rate is limited by the speed of light in copper. With a 2-inch signal path you are limited to something like 3 GHz. The other problem is the one Durinia raised about active interposers and heat. A passive "wiring" interposer is fine, but if you start building things like the memory controller onto the interposer then heat may become a problem. I never got around to posting a response, but basically I think that's a problem that's going to have to be solved sooner rather than later, because we're clearly headed in the direction of 3D assemblies of dies (i.e. stacked HBM2, etc). I know that one proposal is "thermal planes" that act like heatpipes to bring heat out to the sides of the stack, but I don't know the feasibility of that.
|
# ? Jan 23, 2016 21:29 |