![]() |
|
|
#430 | |
|
May 2013
East. Always East.
11×157 Posts |
Quote:
The turbo clocks your hardware up past its design specifications if, and only if, the CPU believes it can survive the heat. Even with stock coolers this is the case. I had an i7-920 which was locked. Its base frequency was 21 x 133 MHz = 2.8 GHz. When all cores were loaded, the Turbo Boost would increase the multiplier to 22 (2933 MHz). When only one core was loaded, the frequency was increased to 3066 MHz. However, the idle cores had to be IDLE. I managed to get a single-core prime stress test to only tax the one core, and I saw the 3.06 GHz sustained for a few minutes (it was rather exciting) but even opening the task manager to check the other cores was enough to de-idle the rest of the processor and the frequency dropped to 2.93 GHz and it stayed there as the task manager itself put enough load on the processor. This all sort of plays into the power saving feature you've mentioned. It's similar, so the confusion is perfectly fine. This feature clocks your processor down (to a minimum of 1.6 GHz, I believe) whenever it's idle. More specifically, it does this on a per core basis. This in tandem with turbo boost is what allows the higher frequencies to be reached. If the processor can run full tilt on four cores, then it must be able to run at least slightly faster on a single loaded core. Unfortunately, there isn't some threshold for bringing the processor up to max speed from 1.6 GHz, so right clicking your desktop is enough of a load that your operating system thinks the processor should ramp up. Ideally, the processor core would only speed itself up if it was under a 10%-20% load. This would be an amazing power saving feature. To be perfectly honest, your average home user and a lot of work users would hardly ever need 3.6 GHz to kick in. My processor runs at some silly voltage like 0.900 V when it's at 1.6 GHz. The efficiency you would see from only running at that kind of speed until you really needed it would be mind blowing. |
|
|
|
|
|
|
#431 |
|
Romulan Interpreter
"name field"
Jun 2011
Thailand
41·251 Posts |
test.
this thread is broken (it doesn't show last page). posts moved? edit: now it is ok, feel free to delete this post. Last fiddled with by LaurV on 2013-11-04 at 06:01 |
|
|
|
|
|
#432 |
|
Jan 2003
2·103 Posts |
Ah, to put in context, we're talking about a Haswell 4770K. Though it could depend on the implementation of the motherboard. My memory is foggy, but I recall from Sandybridge onwards it was advisable to enable turbo. On Asus motherboards (and I'm sure others) you can override the short and long duration power package limits.
The other cores don't need to be idle to achieve an overclock. It can turbo up to a multiplier (e.g. to overclock override the max turbo from 39x to 47x) as long as the power consumption doesn't exceed a level (instead of 84W power limit, override to 255W - which can never be exceeded). In practice, this will make it a permanent overclock when the load from P95 comes. But of course this is only recently, and if I'm not mistaken also requires a K series processor. For Nehalem and earlier it didn't use to work this way. |
|
|
|
|
|
#433 |
|
"Mr. Meeseeks"
Jan 2012
California, USA
42078 Posts |
Thinking of upgrading this old AMD quad to a haswell, and hearing stories of really bad OC's, is it worth getting the K?
|
|
|
|
|
|
#434 |
|
Sep 2002
Database er0rr
5×937 Posts |
IMHO Haswell is great with fast RAM -- it scales well. I would recommend replacing the stock cooler with something special to keep the overclock/overvoltage cool. My 4770k has been up 5 days with the BIOS saying 4.7 GHz (see the previous post of mine in this thread) and now LLR says 4374.22MHz -- this is with less throttling I think due to a lower voltage of 1.27v.
Last fiddled with by paulunderwood on 2013-11-25 at 22:30 |
|
|
|
|
|
#435 |
|
May 2013
East. Always East.
172710 Posts |
Thermal issues will be lesser in the i5 that has no hyperthreading. It has its own graphics. If you have a GPU the iGPU can be turned off... or something. Idle. Whatever. Something like that.
K-chip is still a better chip. Better stability (and probably better thermals) at stock. I'd pay the extra $20. Invest in some good cooling too. H90 or H110, depending on what your case can hold would be nice. |
|
|
|
|
|
#436 |
|
May 2013
East. Always East.
6BF16 Posts |
Get a board with good power delivery. I'd recommend something from ASUS, personally. Get something better than their second or third cheapest. If you're looking to splurge a bit, go for the one with the VRM waterblock for more stable power delivery.
|
|
|
|
|
|
#437 |
|
"Mr. Meeseeks"
Jan 2012
California, USA
37×59 Posts |
Thinking about these two:
MSI Z87-G43 ASRock Z87 Extreme3 Both companies I like, leaning toward the ASrock one, not really willing to spend more than I have to, any Z87 is really just what I need.. As for watercooling, looking at the CM Hyper Evo, it performs more then good on stock, even overclocked. (I don't plan to OC too much) EDIT: Yes, I plan on 2400 MHz memory. Last fiddled with by kracker on 2013-11-25 at 23:55 |
|
|
|
|
|
#438 | |
|
P90 years forever!
Aug 2002
Yeehaw, FL
17×487 Posts |
Quote:
Good luck on achieving stability. My two Haswells are still not stable. I've been boosting voltages and mem timings and last tried OCing mem to only 2133. No luck. Now trying Linux and Win 7 instead of Win 8. |
|
|
|
|
|
|
#439 |
|
Aug 2002
North San Diego Coun
821 Posts |
+1 to Prime95's sentiment.
I have the -45 version (upgraded sound) and wouldn't buy it again either. Won't run 2400 memory at speeds greater than 2133 (but at least it is stable for me 7+ weeks uptime on Win7 x64). |
|
|
|
|
|
#440 | |
|
"Mr. Meeseeks"
Jan 2012
California, USA
1000100001112 Posts |
Quote:
|
|
|
|
|
![]() |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Haswell-E Prelim. Benchmark | sdbardwick | Hardware | 37 | 2015-02-10 18:49 |
| Prime95 and Haswell | Pleco | Information & Answers | 22 | 2014-07-13 16:03 |
| Haswell Rig | TimSorbet | Hardware | 64 | 2014-05-27 13:22 |
| Prime95 version 27.1 early preview, not-even-close-to-beta release | Prime95 | Software | 126 | 2012-02-09 16:17 |
| Missing mouse-over preview text | retina | Forum Feedback | 1 | 2011-09-12 15:32 |