|
Lord Windy posted:I was trying to work out why my computer was crashing on even small overclocks. I also think I found out why my system was running so poo poo before, the fan on my big cooler has seized. Setting VCore ~1.35V should land you stably in the 4.2-4.5 range. Just remember to not disable speedstep - you don't want the CPU to permanently run at 4.2-4.5. In comparison, Auto VCore for my 2500K@4.4Ghz is 1.39V, and it's been happy at that for nearly five years straight. Granted, I've got a Silver Arrow on top of it. BIG HEADLINE fucked around with this message at 07:19 on Oct 28, 2016 |
#
?
Oct 28, 2016 07:17
|
|
|
|
| # ? May 28, 2024 00:25 |
|
|
BIG HEADLINE posted:Setting VCore ~1.35V should land you stably in the 4.2-4.5 range. Just remember to not disable speedstep - you don't want the CPU to permanently run at 4.2-4.5.  I posted in the wrong thread.Thanks though. 4.2 ~ 4.4 would be perfect so I might just go with that.
|
|
#
?
Oct 28, 2016 07:35
|
|
|
What's wrong with leaving the CPU running at knackers out speed? My 3770k is currently at a permanent 4.5Ghz. Am I just using more power than need be, or are there other considerations?
|
|
#
?
Oct 30, 2016 09:49
|
|
|
apropos man posted:What's wrong with leaving the CPU running at knackers out speed? My 3770k is currently at a permanent 4.5Ghz. Am I just using more power than need be, or are there other considerations? Keeping the CPU at max speed all of the time hastens electromigration, which is a fancy way of saying that you're making current/electrons flow too quickly through the transistors, and eventually something has to give. Hint: it's the transistors, not the electrons. That's why chip companies are scratching their heads about what the gently caress they're going to do when they hit the 7-10nm 'wall.' The smaller CPU cores get, the more susceptible they are to electromigration. BIG HEADLINE fucked around with this message at 10:28 on Oct 30, 2016 |
|
#
?
Oct 30, 2016 10:11
|
|
|
apropos man posted:What's wrong with leaving the CPU running at knackers out speed? My 3770k is currently at a permanent 4.5Ghz. Am I just using more power than need be, or are there other considerations? Why would you not let it speedstep back down? You wouldn't be losing performance, but you would save power, and wear on the chip.
|
|
#
?
Oct 30, 2016 10:15
|
|
|
Thanks guys. I just had a play around in my BIOS, loaded optimized defaults and done this: Changed boot order Turned splash screen off Changed core voltage to 1.200 Changed CPU multipler to 45, both in the overall CPU multipler setting and also entered 45 on a 'per core' basis for the maximum multipler. I've rebooted and the thing is still stuck at 4.5GHz in hwinfo and cpu-z, even when I've allowed Windows time to calm down after loading everything. It's the MSI Z77-GD55 motherboard with latest BIOS (1.11). I probably need to take this issue to the overclocking thread to get speedstep working properly.
|
|
#
?
Oct 30, 2016 10:34
|
|
|
VCore of 1.2v is too low for 4.5Ghz. My 2500K is running 4.4Ghz at 1.39v with auto voltage (and SpeedStep enabled).
BIG HEADLINE fucked around with this message at 10:55 on Oct 30, 2016 |
|
#
?
Oct 30, 2016 10:49
|
|
|
BIG HEADLINE posted:VCore of 1.2v is too low for 4.5Ghz. My 2500K is running 4.4Ghz at 1.39v with auto voltage (and SpeedStep enabled). Would that be causing speedstep to stop working, though? It's been nice and steady at 1.200v. I've played about 20 hours of Elite Dangerous with 4.5 G@1.200v and there are no glitches. I know ED isn't the most taxing game though.
|
|
#
?
Oct 30, 2016 10:58
|
|
|
Seems 1.2v is the stock VCore for the 3770K, so never mind. I'm seeing 4.3Ghz@1.2, so you're in the ballpark.
|
|
#
?
Oct 30, 2016 11:05
|
|
|
Ah. OK. I've posted in the overclocking thread so I'll check that later when I get home. I think it's the BIOS on this board not being very user friendly when it comes to setting C states. There's probably a magic combo of settings. It's like a 4 year old board, so I'll forgive it the quirks.
|
|
#
?
Oct 30, 2016 11:16
|
|
|
Enabling C1-C6 and SpeedStep in the BIOS should suffice. Not sure what else would there be.
|
|
#
?
Oct 30, 2016 16:33
|
|
|
Combat Pretzel posted:Enabling C1-C6 and SpeedStep in the BIOS should suffice. Not sure what else would there be. Hmm. It seems to be related to vcore. I've disabled the 1.200v setting and switched it back to auto. My overclock is there, and it's now throttling back down to 1.6Ghz but not for very long: only half a second at a time on each core. It's still 4.5Ghz most of the time. I'm gonna reload defaults and overclock again without changing vcore. Edit: even a very basic overclock without changing vcore works, but it's hardly staying at the lower speed for any decent length of time. Gonna run it at stock 3.5->3.9 :/ apropos man fucked around with this message at 17:36 on Oct 30, 2016 |
|
#
?
Oct 30, 2016 17:28
|
|
|
If you run it at stock, does speedstep work as expected? Ie staying down at a low clockspeed until you actually do something more processor intensive. I'm mostly wondering, have you made sure you don't just have so many background programs running that your cpu is constantly clocking up? I know if I'm just sitting at my desktop with no programs running my cpu tends to stay at the minimum clock, but if I even just have a bunch of browser tabs open and a video playing I see it behaving like you describe, clockspeeds up and down but not staying down for too long. As far as I know it's true that normally you should only have to make sure speedstep is enabled, normally changing voltages shouldn't affect how speedstep works.
|
|
#
?
Oct 30, 2016 19:35
|
|
|
I'm getting normal turbo boost now (3.5->3.9 on a 3770k), so I know it works. When I was trying earlier I was booting into Windows 10 and running hwinfo, CPU-Z and the task manager to get the cpu graphs up. The system has 16GB RAM and a SSD boot drive, so it was barely doing anything at all with one Windows service open and two diagnostic programs running. I also checked advanced power settings, as mentioned earlier. I've given up trying to overclock it for today. I'm guessing that MSI have made it a lot plainer to do on their more modern boards. apropos man fucked around with this message at 19:58 on Oct 30, 2016 |
|
#
?
Oct 30, 2016 19:55
|
|
|
BIG HEADLINE posted:Keeping the CPU at max speed all of the time hastens electromigration, which is a fancy way of saying that you're making current/electrons flow too quickly through the transistors, and eventually something has to give. I don't think there is any solid evidence to support this. As long as your CPU isn't heavily - to - the - max OC'd 24/7 then it will be fine. Yeah your power bills will be higher if it stays at 4ghz all the time but it's not going to kill the chip early.
|
|
#
?
Oct 30, 2016 20:14
|
|
|
Check the Power Options in Windows and see what profile you're running. If it's set to Performance, every little fart a background process does, the clock shoots up. And there's a lot of poo poo in background.
|
|
#
?
Oct 30, 2016 20:25
|
|
|
I was also going to suggest windows. Everything was fine with my overclock till I let windows update and I'm pretty sure it's messed my power settings around. I'm only on a q9550 so the multi can only vary from 6-8.5 but since the last update it's staying pegged to maximum aside from every now and again when it drops down for a second or two. Wasn't like this earlier so I'm pretty sure it's windows, if I figure out which specific setting I'll post it incase it helps. Cheers windows 10
|
|
#
?
Oct 30, 2016 22:15
|
|
|
Lube banjo posted:I don't think there is any solid evidence to support this. As long as your CPU isn't heavily - to - the - max OC'd 24/7 then it will be fine. Yeah your power bills will be higher if it stays at 4ghz all the time but it's not going to kill the chip early. I'm interested in the benefits of disabling the power saving features... I know it's done in competitions and such, but for normal use how does it help?
|
|
#
?
Oct 30, 2016 23:48
|
|
|
GRINDCORE MEGGIDO posted:I'm interested in the benefits of disabling the power saving features... I know it's done in competitions and such, but for normal use how does it help? It's not worth it. Just set your minimum cpu clock to like 40% in the power settings
|
|
#
?
Oct 31, 2016 09:04
|
|
|
AEMINAL posted:It's not worth it. Just set your minimum cpu clock to like 40% in the power settings Def not worth it for me, but why is it a thing?
|
|
#
?
Oct 31, 2016 14:05
|
|
|
GRINDCORE MEGGIDO posted:Def not worth it for me, but why is it a thing? Choices and options are available. It's one of them. For extreme overclocking runs, you probably don't want any of the power saving poo poo in your way, because it might cause some instability switching voltages and frequencies when you're trying to push it to the limit. For everyone's actual normal use, the power saving features are worthwhile.
|
|
#
?
Oct 31, 2016 15:00
|
|
|
All the way up through a QX9650 I was one of those advocates for never using power saving features for extra stability. Starting with Sandybridge (I ended up skipping the 1366 generation) I left everything on because it didn't feel like it was compromising anything anymore. Had a 4.6Ghz Sandybridge and now a 4.7Ghz Haswell. They keep getting better and better about switching states stabily and I think a lot of people who don't use power saving remember the pre Sandybridge era where you were potentially giving up a lot of performance. On the far far end with sub ambient cooling it could be a different story, but for anyone with air or water I think it's mostly lingering thoughts from old power saving working pretty badly than anything based on real world testing of recent hardware.
|
|
#
?
Oct 31, 2016 15:18
|
|
|
apropos man posted:Hmm. It seems to be related to vcore. I've disabled the 1.200v setting and switched it back to auto. I had a motherboard that pegged my CPU at max if I set a specific voltage. Do you have a voltage *offset* mode? It wasn't until I upgraded motherboards to one that had an offset mode that I was able to overclock with a higher voltage and keep speed step at the same time.
|
|
#
?
Oct 31, 2016 15:21
|
|
|
craig588 posted:On the far far end with sub ambient cooling it could be a different story, but for anyone with air or water I think it's mostly lingering thoughts from old power saving working pretty badly than anything based on real world testing of recent hardware. Even then, at this point you're probably talking the difference in 100Mhz of overclockability between stable w/power saving and stable without. Unless you're trying to break records or something, it's just not worth it anymore.
|
|
#
?
Oct 31, 2016 15:30
|
|
|
It looks like we may now have some Kaby Lake desktop chip specs. Not that they're really surprising since like the mobile models we're seeing around a 200-300MHz bump in base clocks across the board, but it remains to be seen what turbo clocks look like and the implications for overclocking.craig588 posted:All the way up through a QX9650 I was one of those advocates for never using power saving features for extra stability. Starting with Sandybridge (I ended up skipping the 1366 generation) I left everything on because it didn't feel like it was compromising anything anymore. Had a 4.6Ghz Sandybridge and now a 4.7Ghz Haswell. They keep getting better and better about switching states stabily and I think a lot of people who don't use power saving remember the pre Sandybridge era where you were potentially giving up a lot of performance. On the far far end with sub ambient cooling it could be a different story, but for anyone with air or water I think it's mostly lingering thoughts from old power saving working pretty badly than anything based on real world testing of recent hardware. From everything I can tell SpeedStep mostly stopped being a liability for overclocking sometime around Nehalem but I'm not really sure since I skipped straight from a Pentium 4/M desktop that doesn't support it at all to an i7-920 that seems to have no issues with it.
|
|
#
?
Nov 1, 2016 00:06
|
|
|
LiquidRain posted:I had a motherboard that pegged my CPU at max if I set a specific voltage. Do you have a voltage *offset* mode? It wasn't until I upgraded motherboards to one that had an offset mode that I was able to overclock with a higher voltage and keep speed step at the same time. When I was overclocking the other day I looked for voltage offset but all of the voltage settings (apart from setting vcore at a specific value, which wasn't exactly obvious) weren't very self explanatory. I was trying to follow another guide I'd seen which mentioned using an offset, just as you do. I dunno, I think it's those older Z77 boards that MSI didn't make a very user friendly BIOS for. Either that or my stupidity. I'm sure the more recent MSI boards are a lot easier. Meanwhile, I'll just run it at stock speeds.
|
|
#
?
Nov 1, 2016 06:51
|
|
|
Power savings and Heat savings for me with my 6 core 3930K. That thing with all the cores pegged will heat my place rather well when Encoding for sure. There is no reason since Sandy Bridge to disable th C states that let the cores Idle to their stock idle frequency (and voltage). However, reasons it may not work are if you set the Voltage manually to a Set Voltage instead of Offset adjustment. (This was the case on 2 SB chips I have overclocked, it may have changed on newer chips/chipsets). If you leave it at auto and see what the voltage gets to when it is at max load stock, then in the bios you want to adjust the Offset up or down (only by .05 to start) and see what it does from there. Usually you want to shoot for 1.35v and adjust from there for stability as you increase the overclock. 1.4v +-.05 is usually more than enough at 4.4-4.6Ghz but a lot of quads can handle 4.4 with 1.3-1.35 (As seen above. 1.2 worked even). Hell my ancient C2Quad Q9550 runs 1ghz over stock at 1.2V at 3.84ghz which usually required 1.3+Vcore on a lot of the same chips back then. You will need to adjust and test to see what you can get and how stable it is at load with testing. There are a lot of options on new boards to dial things in and still save power while overclocking, one thing you will want to take watch of is "Loadline Calibration" which you usually can adjust up and down which is supposed to counter vDroop (the drop in voltage that happen when a large load is placed on the CPU causing the voltage to drop a little) sometimes at more than just the first step up from disabled can be a bit to aggressive and push the voltage up like .2 or .3V which can be enough to cause damage in the long term (or if it is too much, short term). So be sure to test and watch that you aren't shoving 1.5-1.6vcore into the chip with auto or aggressive loadline settings. This is one reason also not to just use the "Motherboards Overclocking Automatic Magic" as usually it just shoves the Voltage up and sets the clock to something like 4.2-4.4Ghz and calls it good, but that is a good way to make sure you are not getting near the optimal performance or longevity you could be with your setup.
|
|
#
?
Nov 1, 2016 17:50
|
|
|
EdEddnEddy, your megapost spurred me into trying again. I get this, still on stock:  Is CPU-Z reporting the correct vcore there, for 3.9GHz?? It can't be that low, surely.
|
|
#
?
Nov 1, 2016 20:40
|
|
|
It turns out that CPU-Z isn't too reliable for vcore. I adjusted to permanent 1.200v and overclocked to 4.5GHz as per the other day. hwinfo showed more accurate information, although not perfectly at 1.200 but that might be inconsistencies in delivery by the motherboard. Anyway, after reading Combat Pretzel's post again: Combat Pretzel posted:Check the Power Options in Windows and see what profile you're running. If it's set to Performance, every little fart a background process does, the clock shoots up. And there's a lot of poo poo in background. I checked power saving options. I had checked it the other day and it was on the 'Balanced' plan, with Minimum Processor State at 5% so I thought that was OK. I decided to lower the Minimum Processor State to 2% and it worked!  The CPU state in CPU-Z and hwinfo is at 1600MHz most of the time now at idle, instead of before when it only dropped down momentarily once every couple of minutes. So the Windows default of 5% may be too high for most enthusiast systems. I'm happy now I've got a usable overclock  4.5GHz on air is pretty sweet. I'm using a Be quiet! Pure Rock cooler, by the way. apropos man fucked around with this message at 21:06 on Nov 1, 2016 |
|
#
?
Nov 1, 2016 21:02
|
|
|
Odd you had to change the 5% to 2%, however each chip has its own sort of lowest clock speed at idle. Mine hits 1.2ghz at idle but 1.6 with probably still lower vcore isn't bad at all. It's nice to help reduce heat/power usage and probably extending the life of your chip immensely rather than those that run their chip at the overclocked state 100% even if there isn't a load applied. Sure the chip may still outlive its usefulness before it dies, but that is always a diceroll in itself. I plan to run this chip into the ground pretty much. Nothing really kills it in single thread process and the only benchmark chip that seems to beat it is any K chip that has 8-10+ cores just in multithread test alone. Also yes, CPU-Z is great to learn about bits of the CPU, but I have been a bigger fan of HWinfo32 to get more accurate sensor readings and it gives you a ton more detail about the entire system to boot.
|
|
#
?
Nov 1, 2016 23:26
|
|
|
I've actually got it running at 1% now. I also plan to keep this system for a while. You can get some bargains going back a gen or two (now three!). I figure that this is an arbitrary scaling by the OS, depending on what it thinks the maximum CPU capability is and won't interfere enough to allow it to crash. Why does Windows need a lower limit, though? I've checked in Ubuntu using /proc/cpuinfo and it's running at 1600MHz fine without any OS intervention to tell it that it can go slow. Bloody Windows. Is hwinfo32 better for sensor accuracy or something? I'm using hwinfo64. apropos man fucked around with this message at 07:03 on Nov 2, 2016 |
|
#
?
Nov 2, 2016 07:00
|
|
|
Kaby Lake-S benchmarks. http://wccftech.com/intel-kaby-lake-core-i5-7600k-review/
|
|
#
?
Nov 3, 2016 17:56
|
|
|
Heh..so Intel are basically re-releasing last year's CPU but replacing the TIM with solder and bumping the clock up 10%. Yikes. That R&D money really must all be going into what happens after "7nm".
|
|
#
?
Nov 3, 2016 19:35
|
|
|
Riflen posted:Heh..so Intel are basically re-releasing last year's CPU but replacing the TIM with solder and bumping the clock up 10%. Yikes. That R&D money really must all be going into what happens after "7nm".  V-- Don't forget that they didn't add solder to the K chips, either, but had the G3258 soldered. Anime Schoolgirl fucked around with this message at 19:54 on Nov 3, 2016 |
|
#
?
Nov 3, 2016 19:47
|
|
|
Riflen posted:Heh..so Intel are basically re-releasing last year's CPU but replacing the TIM with solder and bumping the clock up 10%. Yikes. That R&D money really must all be going into what happens after "7nm". You're saying that like it's something new, this is exactly what they did in 2014 but they didn't even update the GPU then.
|
|
#
?
Nov 3, 2016 19:50
|
|
|
A friend wants to build a desktop for video editing. He was looking at 6800k - will there be a 6 core kaby lake? Also, would kaby vs Skylake make a lot of difference for video editing? How much of a difference would a 4 core i7 vs a 6 core make for video editing?
|
|
#
?
Nov 3, 2016 20:47
|
|
|
The Slack Lagoon posted:A friend wants to build a desktop for video editing. He was looking at 6800k - will there be a 6 core kaby lake? If used hardware is acceptable to your friend, he can get an off-lease HP Z420 or Dell T3600 workstation with a 6-8 core Sandy Bridge-EP in it for under $400. Assuming the editing software he's using is properly multithreaded to take advantage of more cores, that's a killer deal. It may use more power/produce more heat than a newer CPU, but should feel just as fast.
|
|
#
?
Nov 3, 2016 21:01
|
|
|
SamDabbers posted:If used hardware is acceptable to your friend, he can get an off-lease HP Z420 or Dell T3600 workstation with a 6-8 core Sandy Bridge-EP in it for under $400. Assuming the editing software he's using is properly multithreaded to take advantage of more cores, that's a killer deal. It may use more power/produce more heat than a newer CPU, but should feel just as fast. He's using avid Media composer. I'll recommend that but it's not likely he wants to go for that. He is really odd about wanting TOP TIER future proofing. Tired to explain that's not a thing. He originally wanted the 6900 but I shut that down.
|
|
#
?
Nov 3, 2016 21:04
|
|
|
The Slack Lagoon posted:A friend wants to build a desktop for video editing. He was looking at 6800k - will there be a 6 core kaby lake? There won't be a consumer hexacore until Coffee Lake in ~2 years (maybe 18 months) from now. That being said, the cheaper option than going with a Broadwell-E on an X99 board is to get a 6700K, which is a quad core with Hyperthreading. It's still a quad core, but HT simulates another four cores which most higher-quality video editing programs can utilize. Kaby Lake will be ~10% faster than Skylake, but for video the main improvement it'll grant is in *decoding*, not *encoding*. The 6800K will be the faster option, but there *are* reviews out there that will show him if the price difference will be worth it. Someone asked a similar question on Reddit: https://www.reddit.com/r/buildapc/comments/4xos0i/video_editing_cpu_6700k_6800k_or_5820/ BIG HEADLINE fucked around with this message at 21:08 on Nov 3, 2016 |
|
#
?
Nov 3, 2016 21:04
|
|
|
|
| # ? May 28, 2024 00:25 |
|
|
The Slack Lagoon posted:
This one is pretty easy as future proofing is a fools errand at best.
|
|
#
?
Nov 3, 2016 21:09
|
|
