|
necrobobsledder posted:I still think the tablet game is basically over though without a way for developers to port everything over to x86 quickly. Apple may ironically make x86 mobile rather relevant if they release a solid iOS for x86 tool suite something similar to Rosetta in pursuit of their "iOS ALL THE THINGS" strategy that's highly controversial. The big winning move that could be made with x86 tablets requiring minimal cooperation from the Android / Apple side is if a hypervisor could be started that lets users run iOS, Android, Linux, Windows, etc. on their 10" tablets. That's been awful quiet for a while when I last saw Xen making some moves to do it in 2009 for laptops. ARM virtualization is hardly something at the top of vendors' agendas it seems in favor of baking in user profiles instead while x86 virtualization is basically oldhat now. I actually see things going back to the mainframe style of computing, that is all the computational power is on a remote server and that info is beamed back to the tablet via something like RemoteFX. That would be a lot easier than trying to fit a proper hypervisor with all the fixings into a device with limited computational/power resources. You still have the problem of bandwidth doing it everywhere, but I can easily see that working in a LAN environment in the short term. Let the device that's best suited for the task do the work, just make sure everything is on the same page to take advantage of it.
|
#
?
May 31, 2013 18:24
|
|
|
|
| # ? Apr 19, 2024 15:04 |
|
|
From my perspective, it's rather hard to do a lot of development work to build these very things to bring about this revolution (my career is pretty much cloud-cloud-cloud ultrafull stack engineer) on a local semi-fat client (Macbook Air on 4GB of RAM is crappy for running IntelliJ and a bunch of other crap including a Vagrant launched VM). There's stuff like tramp in emacs to edit files remotely pretty transparently and all but a ton of utilities really can't get away from local file access very well (trying to do remote debugging with JBoss from IntelliJ requires a fuckin' local JBoss copy of all things, a weird one I guess). A lot of utilities that work on top of network mounted filesystems tend to assume that they're local basically, and when you're on flakey internet connections, the results can get pretty comical and indeterminate application state is easy to have (granted, that's why we backup and use version control). From the perspective of an enterprise doing VDI to mobile, they get full control over data going in and out of a system, but the cost is enormous and the user experience potentially compromised (lag, buffering, the usual network headaches). Companies do want to give users company-sanctioned devices to do company business on, but mixing their personal and business data is a tricky problem to counter in any environment. It comes down to whether a company is more concerned about data loss / provenance than whether their users are happy with their mobile work environment experience (the primary reason this all took off, right?) and I don't think that the answer is quite obvious. movax posted:Microcenter has some nice discounts on Ivy Bridge...$170 for a 3570K, $230 for a 3770K
|
|
#
?
May 31, 2013 19:51
|
|
|
What was that article that showed only a 5% increase in speed, comparing Ivy and Haswell?
|
|
#
?
Jun 1, 2013 13:17
|
|
|
ijyt posted:What was that article that showed only a 5% increase in speed, comparing Ivy and Haswell? Not the article you're thinking of, and it's only one benchbark, but here, the single threaded score shows the 4770k with 1.74, beating the 3770k which has 1.66, a 5% difference.
|
|
#
?
Jun 1, 2013 15:09
|
|
|
Benchmarks are popping up now. Looks like about the same increase from Sandy to Ivy to me. I'll be keeping my 3570 for a while longer. Big increases in the mobile arena, no surprises there.
|
|
#
?
Jun 1, 2013 15:38
|
|
|
There really isn't any reason for someone to upgrade unless they're on Westmere or old. I didn't think this was unexpected.
|
|
#
?
Jun 1, 2013 15:55
|
|
|
Anandtech's Haswell coverage: Anandtech: The Haswell Review: Core i7 4770K and i5 4560K Tested Anandtech: The Intel Iris Pro 5200 Graphics Review: Core i7 4950HQ Tested Anandtech: Intel's Haswall Quad-core Desktop Lineup Anandtech: Intel's Haswell Quad-core Mobile Lineup
|
|
#
?
Jun 1, 2013 16:22
|
|
|
Well, if you have an unlocked Sandy Bridge or better, there's no real reason to care about Haswell on the desktop. i5-2500K is most definitely the Q6600 of current times. Haswell should make for some compelling laptops though. It should shake up the scene totally
|
|
#
?
Jun 1, 2013 16:32
|
|
|
Ars Technica has a pretty concise overview here. Sounds like it'll be a pretty good upgrade from my Yorkfield 
|
|
#
?
Jun 1, 2013 16:46
|
|
|
I'm on a Phenom II, so it's probably worth the jump to Haswell still, right? I just assume this means there isn't much point to shell out for it if people are on SB/IB.
|
|
#
?
Jun 1, 2013 17:20
|
|
|
Sober posted:I'm on a Phenom II, so it's probably worth the jump to Haswell still, right? I just assume this means there isn't much point to shell out for it if people are on SB/IB. I'd definitely say so.
|
|
#
?
Jun 1, 2013 17:31
|
|
|
So Haswell officially launches June 4th. Is this going to be a hard launch where I'll be able to buy a processor and mobo on the same day? I'm coming from Nehalem so I'm looking forward to seeing this ~40% improvement in performance.
|
|
#
?
Jun 1, 2013 19:13
|
|
|
So Haswell isn't going to overclock any better than Ivy Bridge? That's a bit disappointing.
|
|
#
?
Jun 1, 2013 19:16
|
|
|
I cannot wait to upgrade from my E8400. It's still a great chip and will continue to serve me well, but man, the performance jump should definitely be noticeable.
|
|
#
?
Jun 1, 2013 19:27
|
|
|
Un-l337-Pork posted:I cannot wait to upgrade from my E8400. It's still a great chip and will continue to serve me well, but man, the performance jump should definitely be noticeable. I just went from an E6700 to a e3-1240v2 and man is it hilarious
|
|
#
?
Jun 1, 2013 19:31
|
|
|
Pardon my ignorance but can someone explain which die is which in this Anandtech photo of a Haswell i7-4950HQ: Isn't the Haswell quad core die very rectangular? Does that make the smaller rectangular die to the right the CPU and the massive square die the eDRAM?
|
|
#
?
Jun 1, 2013 19:59
|
|
|
Shaocaholica posted:Pardon my ignorance but can someone explain which die is which in this Anandtech photo of a Haswell i7-4950HQ: Small die is L4 cache according to this:  Multi-chip modules
|
|
#
?
Jun 1, 2013 20:02
|
|
|
Oh I thought all desktop/laptop quad core Intel dies were the exact same layout just different binning and clocks. Wouldn't that make sense from a manufacturing standpoint? Or did Intel do this specifically on the eDRAM Haswells to get it all the fit onto that sized package?
|
|
#
?
Jun 1, 2013 20:08
|
|
|
The GT3 4-core die is huge. 65% of the die area is GPU bits. There were fewer variants in SNB/IVB, yeah. SNB had 4-core HD 3000 (all quad i7s and i5s), 2-core HD 3000, and 2-core HD 2000. Ivy had the same three in HD 4000/2500, plus 4-core HD 2500. Haswell is still pouring out, but so far we have 4-core/4600 and 4-core/Iris Pro, and we'll have 2-core with Iris, 2-core with Iris SoC for Ultrabooks and high-end tablets... Basically having a non-trivial GPU means having more die variants to save area on the ones that don't use the beefiest GPU. Factory Factory fucked around with this message at 20:13 on Jun 1, 2013 |
|
#
?
Jun 1, 2013 20:08
|
|
|
Factory Factory posted:The GT3 4-core die is huge. 65% of the die area is GPU bits. Oh right I forgot Haswell comes with different sized GPU dies depending on the model.
|
|
#
?
Jun 1, 2013 20:10
|
|
|
Are we going to see the quad-core mobile chips with "Iris pro" graphics in 14/15 inch ultrabooks or are they going to only end up in desktop replacements and mobile workstations?
|
|
#
?
Jun 1, 2013 20:24
|
|
|
Naffer posted:Are we going to see the quad-core mobile chips with "Iris pro" graphics in 14/15 inch ultrabooks or are they going to only end up in desktop replacements and mobile workstations?
|
|
#
?
Jun 1, 2013 20:26
|
|
|
Naffer posted:Are we going to see the quad-core mobile chips with "Iris pro" graphics in 14/15 inch ultrabooks or are they going to only end up in desktop replacements and mobile workstations? Isn't the whole point of Iris Pro to replace dGPUs?
|
|
#
?
Jun 1, 2013 20:26
|
|
|
Naffer posted:Are we going to see the quad-core mobile chips with "Iris pro" graphics in 14/15 inch ultrabooks or are they going to only end up in desktop replacements and mobile workstations? There are laptops that are thicker/heavier than ultrabooks and thinner/lighter than desktop replacements that could accommodate an Iris pro part I would think.
|
|
#
?
Jun 1, 2013 20:37
|
|
|
I'm on a q6600, how rocked will my socks be by Haswell?
|
|
#
?
Jun 1, 2013 20:40
|
|
|
^^^^ Hey there Q6660 buddy!   A 3x increase  While Haswell is only a marginal improvement over Ivy Bridge, I just might pull the trigger once things settle down. Although being a cheap bastard, maybe I'll make it until Broadwell  Also, anyone knows what's the deal with the L3 cache?  Was it decoupled for efficiency power efficiency reasons? Because that seems like a pretty big regression in latency right there  The article also doesn't seem to specify which models will support TSX, but that's what I was most looking forward to...
|
|
#
?
Jun 1, 2013 20:56
|
|
|
Fuzzy Pipe Wrench posted:I'm on a q6600, how rocked will my socks be by Haswell? http://www.anandtech.com/bench/Product/836?vs=53 http://i.imgur.com/i3wG2k7.png drat imgur distracts me for 20 minutes every time I try to make an upload
|
|
#
?
Jun 1, 2013 20:57
|
|
|
mobby_6kl posted:^^^^ The -HQ version has the eDRAM cache, which is a fully-integrated L4 cache. On balance performance should be similar, as misses to L3 can be serviced by L4 faster than by system RAM what with higher bandwidth. But yes, it is decoupled into its own frequency domain, as the CPU clearly deprioritizes its power domain when the eDRAM cache is present.
|
|
#
?
Jun 1, 2013 21:04
|
|
|
Anyone know how the eDRAM affects general CPU performance when all other factors are the same? The Anandtech article pits it against a 4770K without eDRAM but that proc is way faster.
|
|
#
?
Jun 1, 2013 21:34
|
|
|
Why did Intel disable TSX on K SKUs?
|
|
#
?
Jun 1, 2013 22:40
|
|
|
Alereon posted:You'll probably see them in 14/15" notebooks but not Ultrabooks. Ultrabooks are thin so use low-TDP processors, these are high-TDP processors. I always read this, but it doesn't explain why I have a laptop that is 1.4kg, 13.1", with discrete Geforce 320M 1GB (23W TDP) and Core i5-520M (2 core, 4 thread, 2.4GHz, boosts to 2.9, 35W TDP). I guess it is 33mm thick (at the thickest point, battery), but it does have a DVD writer (which is unnecessary, save space that way!) It's not like the battery life is somehow totally hopeless or anything (you can switch back to Intel GPU). If I could have full-fat Haswell in a very similar chassis, I would have my next laptop in my sights.. If "Ultrabook" is always going to mean "shaving off a millimetre or two, whilst sacrificing a huge amount of performance", then count me out for that poo poo. HalloKitty fucked around with this message at 22:57 on Jun 1, 2013 |
|
#
?
Jun 1, 2013 22:47
|
|
|
Shaocaholica posted:There are laptops that are thicker/heavier than ultrabooks and thinner/lighter than desktop replacements that could accommodate an Iris pro part I would think.
|
|
#
?
Jun 1, 2013 22:51
|
|
|
HalloKitty posted:If "Ultrabook" is always going to mean "shaving off a millimetre or two, whilst sacrificing a huge amount of performance", then count me out for that poo poo. Well someone just needs to build the thinnest laptop possible around and 'full fat' Haswell.
|
|
#
?
Jun 1, 2013 23:06
|
|
|
Yudo posted:Why did Intel disable TSX on K SKUs? Market segmentation.
|
|
#
?
Jun 1, 2013 23:06
|
|
|
Seems like Intel has resolved the IHS gap issue which was leading people to de-lid their Ivy Bridge CPUs, but Haswell still runs hot as gently caress when you push it to the 4.4-4.5 GHz range (seeing high 60s to high 80s depending on chip/cooler). Seems like the days of "cheap, decent air cooler + change 1 number in bios = 4.5 GHz" are over .
|
|
#
?
Jun 1, 2013 23:06
|
|
|
Nostrum posted:Seems like Intel has resolved the IHS gap issue Link?
|
|
#
?
Jun 1, 2013 23:12
|
|
|
Shaocaholica posted:Well someone just needs to build the thinnest laptop possible around and 'full fat' Haswell. Exactly. It seems like far too many laptops these days are some kind of miserable compromise. We want it all!
|
|
#
?
Jun 1, 2013 23:23
|
|
|
movax posted:Market segmentation. It's loving dumb. I don't mean to direct vitriol at you, but all it does is inhibit adoption by developers--it's not like the install base of Haswell will rival that of the rest of the Core series anytime soon. Perhaps transactional memory will be implemented in other ways; still it's an ISA expansion and should be included. I don't use hypervisors, but leaving out VT-d is likely to piss some people off as well. I would prefer to buy a K SKU, but why should I pay more for a less complete part? It's not like Haswell Ks overclock well enough to cut into the Xeon or whatever market...they all want ECC support anyways. If only I didn't need single thread performance!
|
|
#
?
Jun 1, 2013 23:25
|
|
|
The Xeon E3 lineup is positioned for workstations and doesn't require ECC, there's also LGA-2011, though its current occupants are a bit long in the tooth. I think the basic idea is that they don't want an overclocked i5 or i7 to be a viable option compared to buying more cores. It sucks and I think it's stupid too, but it makes a kind of sense.
|
|
#
?
Jun 1, 2013 23:53
|
|
|
|
| # ? Apr 19, 2024 15:04 |
|
|
I understand market segmentation, but... does anyone even overclock their server CPUs or even workstation CPUs? It seems like the only thing they're achieving is pissing off the SH/SC demographic.
|
|
#
?
Jun 1, 2013 23:59
|
|