|
Agreed posted:Take an already marginal improvement and then cut down how much it means, and really we are in a position of waiting to see if maybe Broadwell will take it further or if it'll be another Tock before we see the kinds of gains that make even the big guys who do pay attention to fractional gains because they mean something on the bottom line perk up their ears.
|
#
?
Jun 7, 2013 01:31
|
|
|
|
| # ? Apr 26, 2024 16:01 |
|
|
The semiconductor industry doesn't really work when modeled with an Econ 101 monopoly. SoCs, CPUs and GPUs are too custom and complex. IMO the better model is a Schumpeterian monopoly: Intel's future business depends on them having a better product that is better and more competitive than its last product, and its dominant market position gives it the resources to pump into R&D and compete with itself that a lower-margin firm cannot bring to bear. If Intel didn't have a better product year after year, 1) it wouldn't be able to sell new PCs to people who already had one, as there wouldn't be an upgrade available, and 2) another company would eventually leverage enough into R&D to dethrone Intel as the market leader. It sounds a little counter-intuitive if all you've heard about monopolies is the Econ 101 version, but Intel being so dominant in the market isn't all bad by a long shot if it means a guaranteed RoI for giant research products that only it can finance at such a speed.
|
|
#
?
Jun 7, 2013 01:43
|
|
|
Factory Factory posted:The semiconductor industry doesn't really work when modeled with an Econ 101 monopoly. SoCs, CPUs and GPUs are too custom and complex. IMO the better model is a Schumpeterian monopoly: Intel's future business depends on them having a better product that is better and more competitive than its last product, and its dominant market position gives it the resources to pump into R&D and compete with itself that a lower-margin firm cannot bring to bear. If Intel didn't have a better product year after year, 1) it wouldn't be able to sell new PCs to people who already had one, as there wouldn't be an upgrade available, and 2) another company would eventually leverage enough into R&D to dethrone Intel as the market leader. That's a really cool point I haven't heard before. Thanks for explaining that; it definitely describes Intel's situation better than my 8th grade understanding of monopolies 
|
|
#
?
Jun 7, 2013 01:49
|
|
|
Factory Factory posted:1) it wouldn't be able to sell new PCs to people who already had one, as there wouldn't be an upgrade available Well, but isn't that basically very much close to what the current situation is? If you have an Intel CPU from the last two years, there's almost zero reason to buy this new one, and, if you follow the roadmap and public statements, not much reason to worry about Broadwell either. Sure, Haswell is better on battery life, and that's good for mobile, but frankly, we're more limited by how much power the screens on these things use, and the current state of battery tech than anything else. From outside of the industry (and more stuck in the 'what do people buy and how do I make money supporting this' view), I'd say that what happened is that Intel no longer views AMD as any sort of threat at all (which is pretty much true) and is scared of ARM eating up all the growth markets, leaving it very little reason to give a poo poo about what desktop/high-power laptop performance is anymore.
|
|
#
?
Jun 7, 2013 02:30
|
|
|
cstine posted:Well, but isn't that basically very much close to what the current situation is? If you have an Intel CPU from the last two years, there's almost zero reason to buy this new one, and, if you follow the roadmap and public statements, not much reason to worry about Broadwell either. Unless you're buying a thermally-constrained device like a laptop or a tablet, form-factors which sell more than desktop PCs and which Haswell is a vast improvement for. Or, for that matter, if you're buying for the enterprise or other high-density computing environment, where yearly power and cooling costs can exceed the cost of new hardware. Factory Factory fucked around with this message at 02:50 on Jun 7, 2013 |
|
#
?
Jun 7, 2013 02:47
|
|
|
Factory Factory posted:Unless you're buying a thermally-constrained device like a laptop or a tablet, form-factors which sell more than desktop PCs and which Haswell is a vast improvement for. Right, but for the tablets/phones wouldn't you be using Silvermont Atoms rather than any of the (unless I missed a few) announced Haswell chips? As for the enterprise market, yeah. I have a few friends who work places like Softlayer and Akamai who would LOVE lower power use/thermals and don't give much of a poo poo if it's faster, as long as they can put twice as many in a rack. I think what's occured here is that Intel has shifted it's attention out of the enthusiast market, because it's so tiny as to not even be a blip compared to cellphones, tablets, and whatever new Macbook apple is selling this quarter. Edit: that, and there's not any real competition in that space anyways. AMD still hasn't got anything competitive, which means the market will result in people buying Intel even if the generational improvements are modest to undetectable. cstine fucked around with this message at 03:10 on Jun 7, 2013 |
|
#
?
Jun 7, 2013 03:07
|
|
|
cstine posted:Right, but for the tablets/phones wouldn't you be using Silvermont Atoms rather than any of the (unless I missed a few) announced Haswell chips? For phones, absolutely yes. For tablets, yes on the low-end. But Haswell has the -Y UULV SKUs for high-end x86 tablets, with ~12-14W TDPs and ~7W SDPs (scenario design power - the system will spike to TDP as temperature allows, especially when switching loads between cores and GPU, but run at the SDP for long-term steady state). Factory Factory fucked around with this message at 03:15 on Jun 7, 2013 |
|
#
?
Jun 7, 2013 03:12
|
|
|
Factory Factory posted:For phones, absolutely yes. For tablets, yes on the low-end. But Haswell has the -Y UULV SKUs for high-end x86 tablets, with ~12-14W TDPs and ~7W SDPs (scenario design power - the system will spike to TDP as temperature allows, especially when switching loads between cores and GPU, but run at the SDP for long-term steady state). I hadn't seen those - any idea when they'll show up on the market? I've kinda had my eye on a proper x86 Win8 tablet, but was going to wait a generation or two, since WinRT is a mess, and the Surface Pro seems an awful lot like a halfway-there product.
|
|
#
?
Jun 7, 2013 03:18
|
|
|
All sorts of crazy poo poo was announced at Computex this week. Here's Engadget's wrap-up article; you can browse it for interesting devices. Related, mmm, Asus Zenbook Infinity  Agreed posted:SH/SC is my happy place. And TFR chat. TFR IRC is great. Is there a SH/SC IRC? I bought mIRC a month or two ago right before Mibbet stopped sucking There's an #shsc on SynIRC, but it's kinda dead. I'll add it to my channel list though. We can start an IRC party. Factory Factory fucked around with this message at 03:53 on Jun 7, 2013 |
|
#
?
Jun 7, 2013 03:22
|
|
|
Factory Factory posted:If Intel didn't have a better product year after year, 1) it wouldn't be able to sell new PCs to people who already had one, as there wouldn't be an upgrade available, and 2) another company would eventually leverage enough into R&D to dethrone Intel as the market leader.
|
|
#
?
Jun 7, 2013 04:14
|
|
|
necrobobsledder posted:I would hesitate to call PCs the same sort of "durable goods" economic category as refrigerators, cars, and construction equipment, but I was moreso looking at the angle from the perspective that Intel can only push the rest of the market (and more importantly, their customers) around so much, including their DRAM partners when their customers' shiny object dreams are changing in directions that put Intel in a position where they're not that important for the solution. They actually kind of are. Sure they haven't reached the point like cars where the average one on the streets is 10 years old yet, but there's tons of people using them for 4+ years, and will be a lot more as we proceed onwards.
|
|
#
?
Jun 7, 2013 04:18
|
|
|
Validation engineer here. The tests for stability that overclockers run may get lucky and actually exercise some speedpaths for certain operations but you are probably not hitting the speedpaths of the whole chip unless you exercise all the features, which you really can't do without intimate knowledge of the platform and how to program it. You need to know what the actual slowest paths in the chip are and you need directed tests for those paths and you need to run them with the chip heated up way above normal temperatures or cooled to freezing and and then you've got a good place to start for speed binning. Without knowing the actual worst paths through the chip (which is a job for a whole department by itself) you'll run into things where changing the screen resolution or a cache miss or putting the mouse in the wrong place on the screen triggers an actual slower path and it crashes for no apparent reason. The space is just to big to cover without directed tests. If you could do it with any amount of luck and randomization I wouldn't have a job. EIDE Van Hagar fucked around with this message at 05:55 on Jun 7, 2013 |
|
#
?
Jun 7, 2013 05:31
|
|
|
Factory Factory posted:The semiconductor industry doesn't really work when modeled with an Econ 101 monopoly. SoCs, CPUs and GPUs are too custom and complex. IMO the better model is a Schumpeterian monopoly: Intel's future business depends on them having a better product that is better and more competitive than its last product, and its dominant market position gives it the resources to pump into R&D and compete with itself that a lower-margin firm cannot bring to bear. If Intel didn't have a better product year after year, 1) it wouldn't be able to sell new PCs to people who already had one, as there wouldn't be an upgrade available, and 2) another company would eventually leverage enough into R&D to dethrone Intel as the market leader. 2 doesn't follow from the rest of your post. How does a company that makes massive profits from its monopoly get out R and Ded by some no name. It's my understanding that you can't just genius your way to an amazing chip. You need a shitload of engineering expertise some no name ain't gonna have. I understand your disdain for those Econ 101 types but sometimes a monopoly is just a monopoly.
|
|
#
?
Jun 7, 2013 06:34
|
|
|
The upgrade to intel is going to be big for me but just curious because the reception to Haswell is rather lukewarm, are the temps between it and IB vastly any different at idle/load? I heard Haswell isn't doing so well with OC'ing, hitting really high temps at 4.4GHz+. I'm more looking to do something a bit more modest in the OC department and just using an air cooling solution, so is there any reason to pick up Haswell over IB? Could someone summarize for me what's new about the features on the LGA1150 mobos as well? Are they worth picking over IB? And is that USB3.0 bug that big a deal? Sounds like you just might need to occasionally disconnect/reconnect any storage devices or something?
|
|
#
?
Jun 7, 2013 06:37
|
|
|
Omelette du Fromage posted:2 doesn't follow from the rest of your post. How does a company that makes massive profits from its monopoly get out R and Ded by some no name. Let me expand: 2: If Intel WERE an Econ 101-style monopoly, then the logic goes that it would have no incentive to invest in new products, because it could keep selling the same old poo poo at crazy-high prices. As such, a company with lesser resources but with the will to innovate and invest in R&D could slowly but surely begin to compete again and disrupt Intel's monopoly. Therefore, Intel DOES have an incentive to invest in new products, and this incentive comes from the threat of competition. Therefore, the kind of market dynamics suggested by the kind of monopoly you learn about in Econ 101 doesn't seem to model reality very well. Joseph Schumpeter's model of monopoly, in which a dominant firm will innovate in order to maintain its dominant market position, using the resources its dominant position grants it, fits better. Intel is still pricing above the market's competitive equilibrium, like a monopoly would, we're just talking about the dynamics beyond a simply supply and demand curve for interchangeable widgets. quote:I understand your disdain for those Econ 101 types but sometimes a monopoly is just a monopoly. I'm not disdainful. Econ 101 monopoly sounded better than "a monopoly, like, y'know, a regular monopoly monopoly." All the Econ 101 concepts remain crazily useful and relevant in further studies.
|
|
#
?
Jun 7, 2013 06:56
|
|
|
Sober posted:And is that USB3.0 bug that big a deal? Sounds like you just might need to occasionally disconnect/reconnect any storage devices or something? I don't think I've ever even had a situation where I was using a file on a USB device that I needed to access again after sleeping the PC. I'm assuming the bug is only at the OS level right, and won't affect installing Windows from USB?
|
|
#
?
Jun 7, 2013 07:32
|
|
|
Shaocaholica posted:I was speaking for the mentality, not the actuality. If you wrote something that would crash an otherwise 'stable' CPU and released that into the overclocking community, they would reevaluate their view of stable. kwinkles posted:Without knowing the actual worst paths through the chip (which is a job for a whole department by itself) Omelette du Fromage posted:2 doesn't follow from the rest of your post. How does a company that makes massive profits from its monopoly get out R and Ded by some no name. JawnV6 fucked around with this message at 07:40 on Jun 7, 2013 |
|
#
?
Jun 7, 2013 07:38
|
|
|
JawnV6 posted:In this case, it'll come from the ARM ecosystem, a diverse set of companies competing for every facet of the design process. They're at the low end of power/performance now, but history is littered with big high-end companies that had the low end of their markets taken away by companies who learned while doing it and worked their way up the value chain. I like Microsoft's strategy re: Win8 and WinRT in this respect. They're forcing Intel to be honest and not rest on their laurels in terms of Windows low power performance while still getting a cut of the action. WinRT may suck, but at least it's forcing out better and better iterations of x-Trail Atom processors on the x86 side.
|
|
#
?
Jun 7, 2013 08:46
|
|
|
Factory Factory posted:Let me expand: Haswell is primarily a mobile focused CPU. Do you agree? It does not do very much for desktop users - where Intel already has the dominant position. I don't want to say that Intel is ignoring desktop users for mobile users (where it isn't as dominant), but the word 'ignore' isn't as far off as it should be. I understood what you meant when you discussed Schumpeter's model, I just didn't buy your reasoning. It seems to be that Intel is only making substantial gains in the areas where it doesn't already have a commanding lead. fake edit: I would argue that Haswell is basically selling the same old poo poo at crazy-high prices.
|
|
#
?
Jun 7, 2013 12:35
|
|
|
How again is Intel 'ignoring' desktop users
|
|
#
?
Jun 7, 2013 14:19
|
|
|
I think Omelette du Fromage is arguing that Intel appears to be skimping out on motivating desktop users to upgrade, due to certain design decisions with Haswell that some of them would have wanted. Though perhaps this "some of them" is just the enthusiasts who notice minute things like the IHS gap on the Haswell series, or the TIM used under said IHS.
|
|
#
?
Jun 7, 2013 14:49
|
|
|
I'm just not sure how much further you can innovate the desktop space based on what people use their computers for. For the past 2-3 year, the most substantial upgrade you could do for general computing was install an SSD, not upgrade a processor or graphics card. Even in enthusiasts brackets, games have been limited by consoles for the most part. It's only in business where any of this extra speed is necessary and intel's been doing fine there. Also, it's easy to dismiss Haswell improvements as being "mobile" only, but they are also going to be huge in the server space where every bit of density counts. So, intel is focusing on a market where they have complete dominance right now. bull3964 fucked around with this message at 15:11 on Jun 7, 2013 |
|
#
?
Jun 7, 2013 15:08
|
|
|
bull3964 posted:I'm just not sure how much further you can innovate the desktop space based on what people use their computers for. For the past 2-3 year, the most substantial upgrade you could do for general computing was install an SSD, not upgrade a processor or graphics card. I think most of the next few years is going to result in much better graphical fidelity in games - but due to nothing involving the CPU. GPUs have gotten remarkably fast, and since they're now putting 4 and 6 and 8gb of ram on the cards (and, combined with the fact that you're not stuck in 256mb-land on the next gen of consoles - especially not the PS4) you're going to see games using bigger and bigger textures which directly relate to games looking loving spectacular - look at Bioshock Infinite, which will use as much vram as you can throw at it. Part of what I'm bemused about is that for the first time in the last 10 or 15 years, I'm not going to be on a 1 or 2 year upgrade cycle - more likely this is going to be a 5+ year cycle, since there's no reason to worry about upgrades (for a desktop) until MAYBE Skylake, maybe.
|
|
#
?
Jun 7, 2013 15:25
|
|
|
Reading up on all the haswell overclocks, I'm glad I went ahead and just bought the 3570k. I'm running perfectly stable at 4.6ghz.
|
|
#
?
Jun 7, 2013 16:46
|
|
|
I'm just hoping I can leapfrog from Sandy Bridge to Skylake... That would be such an ideal scenario and seems easy enough.
|
|
#
?
Jun 7, 2013 17:46
|
|
|
Sober posted:so is there any reason to pick up Haswell over IB? Haswell is better than IB, just by margins that are disappointing to people. If you're buying something right now, the only reason to buy an IB is if you can get it significantly cheaper than Haswell (which won't happen at retail prices, but could happen with a good inventory clearing sale). Tab8715 posted:I'm just hoping I can leapfrog from Sandy Bridge to Skylake... That would be such an ideal scenario and seems easy enough. Haswell looks pretty tempting to me, and I'm on a Nehalem right now. Intel's been doing a steady 15-20% performance improvement each new architecture, which I think makes for a compelling upgrade after 3 of them.
|
|
#
?
Jun 7, 2013 18:15
|
|
|
Looks like ASUS' ROG line just went up on NewEgg http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007627%2050001315%20600438202&IsNodeId=1&name=ASUS&Order=PRICED&Pagesize=20 Edit: I swear they were just there... Links: New-ShitPost fucked around with this message at 18:22 on Jun 7, 2013 |
|
#
?
Jun 7, 2013 18:17
|
|
|
kwinkles posted:
What about pseudo-random?!
|
|
#
?
Jun 7, 2013 18:18
|
|
|
Zhentar posted:Haswell is better than IB, just by margins that are disappointing to people. If you're buying something right now, the only reason to buy an IB is if you can get it significantly cheaper than Haswell (which won't happen at retail prices, but could happen with a good inventory clearing sale). If you buy both a board and chip from Microcenter the IB is going to be a hundred or more cheaper than Haswell though.
|
|
#
?
Jun 7, 2013 18:21
|
|
|
You do get more native USB3 SS and SATAIII ports with a Haswell board vs. an Ivybridge
|
|
#
?
Jun 7, 2013 18:26
|
|
|
But IB already had plenty, so that doesn't really matter.
|
|
#
?
Jun 7, 2013 19:45
|
|
|
Zhentar posted:But IB already had plenty, so that doesn't really matter. IB had 2 SATA III ports, which is fine for me but some people want or need more I'm not saying it's a big deal, but it's just something else to include in the price difference between IVB besides just a small performance bump WhyteRyce fucked around with this message at 20:05 on Jun 7, 2013 |
|
#
?
Jun 7, 2013 20:02
|
|
|
compwhizii posted:Looks like ASUS' ROG line just went up on NewEgg Nowhere to be found on their Canadian site.  Hopefully either they or NCIX stock them soon.
|
|
#
?
Jun 7, 2013 22:59
|
|
|
A few observations on the i7 4770k I received: Coming from an i5 760, Haswell is much faster even in everyday applications like web browsing. For some of the simulation stuff I run, HT is fantastic; for less optimized shitware the single threaded performance is also noticeably better. If you are running a Core2, an SSD is still a better upgrade but a new CPU does make a difference. AES-NI is baller. It runs hot (28 idle, 60 load), but AVX is what does it. Just FPU heavy programs like Prime95 are pretty tame temperature wise even with a 20% OC. Enabling AVX linpack on OCCT sends temperatures skyrocketing. My chip seems to only have modest overclocking potential. If you want to OC, don't skimp on the cooler. Intel's SATA controller is a million times better than the Marvel crap on my old board. Intel's network controller is also great to use. Asus's UEFI is pretty well laid out overall making the transition from old timey BIOS easy. I did turn off fast boot as it is a bit too fast. I have mixed feelings about Asus AI Suite, but Asus's fan configuration is awesome. It makes me want to replace everything with PWM fans. It's hot, fast, and perhaps a worthwhile upgrade if you are still on Penryn or Nehalem. However, Haswell is not so compelling to make waiting to see what AMD's next APU shapes up as a bad idea.
|
|
#
?
Jun 8, 2013 05:08
|
|
|
28 idle is hot? My Q9450 idles at 50.
|
|
#
?
Jun 8, 2013 08:26
|
|
|
ijyt posted:28 idle is hot? My Q9450 idles at 50. 28c idle is fine--though keep in mind Haswell has sleep states that older Core CPU's do not: idling it's using only 10 watts. 60 under load is pretty hot considering I'm using a big heat sink (though perhaps with a bit too much TIM). This gets up to 70 using AVX enabled linpack/prime95 small FFT, a conservative CPU fan profile @ 3.7 ghz. Increasing the voltage makes the problem all the more dramatic. I'm not really interested in pushing it; it's pretty flexible voltage wise under 4.2 ghz. My only contrast is a Nehalem i5, which out of the box could add 700 mhz with little to no additional heat. At stock it's a very fast CPU, it would just be nice if it were less temperamental to OC. Yudo fucked around with this message at 10:00 on Jun 8, 2013 |
|
#
?
Jun 8, 2013 08:46
|
|
|
ijyt posted:28 idle is hot? My Q9450 idles at 50. Do you have a very aggressive overclock going? If not, maybe you should reseat your heatsink because that sounds way too high. Every Intel processor I've owned in the core 2 era and beyond has idled around 30 degrees Celsius.
|
|
#
?
Jun 8, 2013 14:40
|
|
|
If anyone is curious: After 4.4 ghz voltage requirements start to increase quickly. 4.5 @ ~1.23 volts seems to work tested with Prime95 blend and large FFT, but I can't dissipate enough heat to use the cache only small FFT, which really loads the CPU as it does not use system memory. Lower voltage yields a quick BSOD--there is a big gap between what will work at 4.4 and 4.5. Judging by cursory testing, below 4.5 ghz, a high end air cooler can keep temperatures under 90c. Above that my particular chip may be stable but I can't test it properly. How to keep Haswell cool at 1.3 volts as seen by those who get above 4.5-4.6 ghz is beyond me. This was quick and dirty; C states, HT and Turbo were all enabled.
|
|
#
?
Jun 8, 2013 19:16
|
|
|
Yudo posted:How to keep Haswell cool at 1.3 volts as seen by those who get above 4.5-4.6 ghz is beyond me. Haven't you come across any of the posts ITT about popping the IHS?
|
|
#
?
Jun 8, 2013 19:29
|
|
|
|
| # ? Apr 26, 2024 16:01 |
|
|
Shaocaholica posted:Haven't you come across any of the posts ITT about popping the IHS?
|
|
#
?
Jun 8, 2013 19:46
|
|
