EPT13% is extremely adventurous, it was looking more like 4-10 (depending on application) IIRC
Computer Build Resource Thread - Page 1511
| Forum Index > Tech Support |
When using this resource, please read FragKrag's opening post. The Tech Support forum regulars have helped create countless of desktop systems without any compensation. The least you can do is provide all of the information required for them to help you properly. | ||
|
Cyro
United Kingdom20323 Posts
13% is extremely adventurous, it was looking more like 4-10 (depending on application) IIRC | ||
|
Myrmidon
United States9452 Posts
Something that can heavily exploit AVX2 could benefit more. Also, some multi-threaded code with significant dependencies could benefit from TSX in the future. That said, K edition processors are not getting TSX support, for whatever reason. But depending on what they're really locking down, non-K processors might be able to overclock a lot...? Rumors are also that temperatures are still high; overclocking probably not a whole lot better. | ||
|
Craton
United States17275 Posts
| ||
|
Alryk
United States2718 Posts
| ||
|
Craton
United States17275 Posts
| ||
|
Cyro
United Kingdom20323 Posts
Aparantly NDA lifting soon. before anyone cries - Intel told us quite clearly the NDA is at 12:01am June 2, you can pick any time zone in the world. We picked NZ Hopefully this means some more informed information about OC-ing. | ||
|
Cyro
United Kingdom20323 Posts
First thoughts, looks ok. A little suprising in some places. ![[image loading]](http://images.anandtech.com/graphs/graph7003/55326.png) ![[image loading]](http://www.pcper.com/files/review/2013-05-30/x264-2_0.png) Overclocked performance is all over the place. Seen everything from 1.44v 4.5 to 1.3v 4.7 so far (edit; much lower seen, 1.15 4.5/4.6 a few times), and not seen many numbers. Looks like ivy but slightly hotter*, i didn't expect as much as 12-14% gains in x264 though. *no idea how it performs with delid yet Our first-hand information involves a high double-digit number of processors, including samples and final shipping boxed CPUs. Sort testing was limited to 1.2 V to keep heat manageable. Ring/cache ratios are pegged at 3.9 GHz, with the memory controller operating at 1,333 MT/s. Of the chips available for sorting, only one is stable at 4.6 GHz under full load. A few are capable of operating at 4.5 GHz. More run stably at 4.4 GHz. Most are solid at 4.3 GHz and down. Yep, looks like ivy. Time to get out the vice and hammer :D Solid info. 70% of CPUs can clock to 4.5GHz 30% of CPUs can clock to 4.6GHz 20% of CPUs can clock to 4.7GHz 10% of CPUs can to 4.8GHz Overall you will find most CPUs capable of reaching 44x to 45x with varying levels of voltage. These ASUS results were obtained with sealed water cooling systems that are comparable to a Corsair H80/H100 configuration or extremely efficient air cooling with 120mm push/pull fans while applying a maximum core voltage of 1.275v under full thread load conditions. 1.275v is nothing. I think Haswell will be fine on 1.4v delidded which gives a good shot at 5ghz, but it'll beat ivy for x264 as low as 4.4 (from what i have seen from quite a few benches) even if higher performance margins in a lot of games are not nearly that big (it's very hard to say because aparantly nobody knows what a CPU bound game is) ![[image loading]](http://images.hardwarecanucks.com/image//skymtl/CPU/INTEL-HASWELL/INTEL-HASWELL-303.jpg) | ||
|
Ropid
Germany3557 Posts
| ||
|
Myrmidon
United States9452 Posts
AnandTech results with HD 5.0.1. What's new in version 5.0.1? [...]We also updated the benchmark with revision 2200 of the x264 encoder http://www.techarp.com/showarticle.aspx?artno=520 commit b924133cabd125286488e16cfa71488ad4105d63 r2243 Author: Jason Garrett-Glaser <jason@x264.com> Date: Mon Nov 12 10:28:53 2012 -0800 AVX2/FMA3 version of mbtree_propagate First AVX2 function for testing. Bump yasm version to 1.2.0 for AVX2 support. http://x264.nl/x264/changelog.txt <- search for instances of AVX2 Should be more than 12-14% with more recent builds, in the future. [emphasis added all over the place] | ||
|
Cyro
United Kingdom20323 Posts
It kind of makes sense that it isn't that different from Ivy for overclocking. It uses the same 22 nm process and those 3-D transistors, right? We had at least a batch or a few batches of ivies that did 4.5ghz on 1.05v - there was speculation it was due to improvements or testing in manufacturing process that would be standard on haswell. Not the case, but it would have been too good to be true anyway. @Myrm post; As soon as we get AVX2 benches, everything else should look silly next to haswell. It's not the leap i hoped for in terms of general singlethreaded performance, temperatures (ivy runs ~22c cooler with delid - if haswell ran at 65c on a clockspeed+voltage that would put an unmodded ivy at 90c while also having 5-10% IPC gains, it would be really great - but turns out to be unrealistic desires; hopefully it will see similar massive temp drops with delidding) but it looks great just for the leaps in some areas and avx2 support | ||
|
Myrmidon
United States9452 Posts
| ||
|
Ropid
Germany3557 Posts
I was hoping the integrated VRM could mean lower voltage when overclocking. I was thinking this because there was a very large difference in voltage needed between simply passing some short stress test and actually getting the whole CPU running stable in daily tasks. Those tests using Intel's linpack could for example pass with something like 1.21 V for me, but the PC was only completely stable with something between 1.23 and 1.25 V. That 0.04 V difference was what I was thinking the integrated VRM could fix. | ||
|
Cyro
United Kingdom20323 Posts
| ||
|
Ropid
Germany3557 Posts
| ||
|
Myrmidon
United States9452 Posts
From a chip designer's standpoint, all the features and engineering that went into the FIVR, the extra options exposed with respect to BCLK bootstraps, other tweaks, etc.—these all seem like a lot of things to play with and a lot of improvements. These types of features and design that are implemented are the kinds of things they would think about. In fact, most staff aren't involved with testing these things to find the actual limits, particularly not beyond stock conditions. Would they even really know how well these things overclock in shipping conditions? I mean, even the ones testing chips are mostly dealing with internal revisions, engineering samples, most of the way, that are all way under release clock speeds anyway. For the performance-oriented enthusiast, it's maybe more the actual results and clock speeds achievable that are of interest. And it's not particularly dreamy in this respect. | ||
|
Alryk
United States2718 Posts
In y'alls infinite knowledge and collective expert opinions, does this seem worth it for the ~100$ discount you could get with a 3570k from microcenter? (CPU price + mobo discount - tax) Does it also seem like current ivy bridge parts would overclock better than brand new haswell parts? I'll probably end up with Haswell anyways, since I don't think I'll be going much over 4.7ghz anyways.. | ||
Antoine
United States7481 Posts
On June 02 2013 01:04 Alryk wrote: Hmm. In y'alls infinite knowledge and collective expert opinions, does this seem worth it for the ~100$ discount you could get with a 3570k from microcenter? (CPU price + mobo discount - tax) Does it also seem like current ivy bridge parts would overclock better than brand new haswell parts? I'll probably end up with Haswell anyways, since I don't think I'll be going much over 4.7ghz anyways.. with Microcenter's 15 day return policy, if you're even considering buying a 3570k and are nearby a microcenter you should probably go buy one now, since that deal ends monday. if you change your mind you can just go return, if price drops they'll give you the difference, and if you don't buy it now, they don't have a comparable deal, and you end up deciding to go for the 3570k, you'll regret it | ||
|
Cyro
United Kingdom20323 Posts
My understanding is a h100i can take very close to 1.4v on ivy or over 1.5v delidded | ||
|
Cyro
United Kingdom20323 Posts
after, 1.33v, max temp 80c. not known cooling yet, but that's solid >=25c drop on same voltage and temps, assuming the 100c limits were hardly touched before | ||
|
skyR
Canada13817 Posts
| ||
| ||
