|
Spiritus Nox posted:So, just to be clear, that's far beyond the safe limit? Is there any safe amount of OC at all? In a laptop, there really isn't any room for a 24/7 safe OC. You can get a performance boost if you're willing to risk damage that will lower the lifespan of the system.
|
# ? Nov 24, 2012 02:53 |
|
|
# ? May 15, 2024 16:13 |
|
Factory Factory posted:It mostly doesn't, you're right. You can adjust the BCLK and RAM settings (BIOS setup -> Advanced tab -> Performance Options Menu, then enable the overclocking feature), but that won't get you very far. Instead, check out ThrottleStop to increase the power/heat budget for your CPU and let it run at a higher speed. But for God's sake, WATCH THE HEAT. All right, so the impression I'm getting is that I'm not going to be able to get the kind of overclocking I'd like without running a serious risk of bricking my laptop (which I do NOT want to do) - is that right?
|
# ? Nov 24, 2012 02:53 |
|
Not "bricking," per se, as long as you're careful with heat management and don't make any changes permanent. But the system might kick the bucket sooner than it otherwise would due to the extra heat causing wear and tear to pretty much every component. A silly example: if you ran the system pushed far enough to overheat, and you closed the screen, the heat might discolor the screen. For the record, when we say "safe" overclock, that means 100% guaranteed to last the warranty period of the parts (i.e. 3-5 years, depending) even if it runs at 100% utilization 24 hours a day, 7 days a week. That's not a particularly realistic workload; if you're willing to compromise on the system lasting forever, there's room above and beyond that.
|
# ? Nov 24, 2012 02:56 |
|
Factory Factory posted:Not "bricking," per se, as long as you're careful with heat management and don't make any changes permanent. But the system might kick the bucket sooner than it otherwise would due to the extra heat causing wear and tear to pretty much every component. I see. I feel like I'm being a bit dense here, so thanks for bearing with me for a few more questions: If I want to try and get 3 GHZ out of this CPU, will I be able to do so without risking serious and immediate damage? Would you have any idea how much I'd be shortening my computer's life span by overclocking to that degree? And if I went back in the BIOS and dialed things back down at the end of a gaming session that demanded the extra CPU muscle, would that make a significant difference over having it OCed all the time? And lastly, is there anything to heat management other than "Don't push your computer too long for too long or keep it in hot spots?" I know I'm dumping a lot of questions on you here - thanks a lot for the help! I'm in uncharted territory here and trying to weigh the risks.
|
# ? Nov 24, 2012 03:01 |
|
Turbo boost raises the frequency to different amounts under load. Chances are, you aren't going to get a sustained 3 GHz on all four cores no matter what, since the default behavior is 2 GHz/2.6 GHz with four or three cores active/2.8 GHz with two cores active/2.9 GHz with one core active. You can't easily make up that 400 MHz on a laptop. I can't find a full performance scaling review of Planetside 2, so I don't know if 1) you're at 2 GHz and Throttlestop would kick you to 2.6 GHz and solve a lot of problems, or 2) you're already at 2.6 GHz and Planetside 2 is just a ridiculously intensive game. If it's the latter case, there's not much you can do. You might want to look into lowering the terrain, particle, and effects quality, as those can effect the game's CPU usage.
|
# ? Nov 24, 2012 03:26 |
|
Factory Factory posted:Turbo boost raises the frequency to different amounts under load. Chances are, you aren't going to get a sustained 3 GHz on all four cores no matter what, since the default behavior is 2 GHz/2.6 GHz with four or three cores active/2.8 GHz with two cores active/2.9 GHz with one core active. You can't easily make up that 400 MHz on a laptop. Well, the minimum required CPU is 3 GHZ and the recommended is 3.3, and my performance is okay until I get into a firefight with dozens of dudes around, so I would guess we're looking at the latter scenario. I'll try playing with the setting and seeing if I can get it to something playable. So the safe bet here would be to assume I'm not getting significantly more out of my CPU without melting through my computer, yes?
|
# ? Nov 24, 2012 03:33 |
|
You can't compare a 3 GHz Core 2 Duo to the clock speed of a Sandy Bridge CPU. The equivalent real-world performance on Sandy Bridge is about 2.2-2.3 GHz. Your hurdle isn't as high as you think. But you may need to adjust your expectations re: what the "minimum" in "minimum requirements" means.
|
# ? Nov 24, 2012 04:04 |
|
Factory Factory posted:You can't compare a 3 GHz Core 2 Duo to the clock speed of a Sandy Bridge CPU. The equivalent real-world performance on Sandy Bridge is about 2.2-2.3 GHz. Your hurdle isn't as high as you think. But you may need to adjust your expectations re: what the "minimum" in "minimum requirements" means. Oh - Oh duh. The 'dual core' completely flew over my head when I was reading the system reqs. Well now I'm completely paralyzed with indecision. Last request and then I'll stop pestering you for the night. What do you think - give Throttle Stop a shot and see if I can get a good performance/heat balance, or do you think I'd be taking too much of a risk?
|
# ? Nov 24, 2012 04:19 |
|
Get a CPU temperature monitoring program (HWiNFO64 or CoreTemp) and watch it yourself, both before and after enabling Throttlestop. Below 72 C and you have temperature to spare. Above 72 C, you're causing slow, long-term heat damage. Above 90 C, you're potentially causing damage fast enough to effect the health of the system within months, if not sooner. At 100 C, the CPU will throttle itself to prevent immediate damage. Use Prime95 to put a load on the CPU while you play around with settings.
|
# ? Nov 24, 2012 04:26 |
|
Factory Factory posted:Get a CPU temperature monitoring program (HWiNFO64 or CoreTemp) and watch it yourself, both before and after enabling Throttlestop. Below 72 C and you have temperature to spare. Above 72 C, you're causing slow, long-term heat damage. Above 90 C, you're potentially causing damage fast enough to effect the health of the system within months, if not sooner. At 100 C, the CPU will throttle itself to prevent immediate damage. Excellent, I'll do that. Thanks a lot!
|
# ? Nov 24, 2012 04:28 |
|
Well, all ThrottleStop seems to let me do is adjust the CPU multiplier - and when I kick it up one step, running Prime95 shoots my temperatures skyward and I need to kill it before my computer shuts itself down. Is that a sign that it's time to let this one go?
|
# ? Nov 24, 2012 06:09 |
|
Kick it from what to what?
|
# ? Nov 24, 2012 06:23 |
|
Default is 16, I took it to 17, ran prime's 'blend' torture test, frantically shut it off as temperatures shot from sub-60 to over 90. To be fair, even at 16 the torture test made my temperatures crawl over 80.
|
# ? Nov 24, 2012 06:26 |
|
Spiritus Nox posted:Default is 16, I took it to 17, ran prime's 'blend' torture test, frantically shut it off as temperatures shot from sub-60 to over 90. To be fair, even at 16 the torture test made my temperatures crawl over 80. Betcha that if you watch HWiNFO64, the multiplier/frequency doesn't match Throttlestop's setting. 80 to 90 C is, sadly, a common load temperature for many laptops, and the M17x is no exception. This is what we meant by "no thermal headroom," "bad idea," etc. You just can't build a laptop to sustain a load the way you can even with something the size of an Xbox 360. That confirms that there's not much you can do within the realm of reason. Sorry.
|
# ? Nov 24, 2012 06:40 |
|
Factory Factory posted:Betcha that if you watch HWiNFO64, the multiplier/frequency doesn't match Throttlestop's setting. Fair enough. That'll teach me to just get a desktop next time I get a gaming rig. Hopefully I can do something with the settings to reduce the strain on my CPU, it's really only a problem in one game. Thanks for the attempt!
|
# ? Nov 24, 2012 06:44 |
|
I'm trying to overclock my AMD Phenom X6 1100T and I'm wondering if anyone has any experience with this particular model. I want to get as much juice out of it as possible (the 20x multiplier taking it from 3.3GHz to 4.0GHz seems sufficient), but am uncertain of what voltage rating would be appropriate for that clock. And yes, I have sufficient aftermarket cooling and a heavily cooled case to support high temperatures.
|
# ? Nov 25, 2012 08:31 |
|
Deutsch Nozzle posted:I'm trying to overclock my AMD Phenom X6 1100T and I'm wondering if anyone has any experience with this particular model. I want to get as much juice out of it as possible (the 20x multiplier taking it from 3.3GHz to 4.0GHz seems sufficient), but am uncertain of what voltage rating would be appropriate for that clock. Go ahead and try it with a 20x multiplier then, this is a black edition Phenom which means all you really need to do is just raise the multi. You will probably need a voltage above 1.4v. Do not pass 1.55 if you're air cooled. If it boots then run prime95 for a couple of hours and make sure there's no errors or workers dropping. Most of the time you BSOD if you need more voltage or a lower clock speed. Watch temperatures, do not let it go past 60c while stressing.
|
# ? Nov 25, 2012 12:33 |
|
An Unoriginal Name posted:Go ahead and try it with a 20x multiplier then, this is a black edition Phenom which means all you really need to do is just raise the multi. That doesn't sound right to me. He's overclocking a Phenom 2, which is made from a 45nm process. Doesn't that give him more leeway for high thermals compared to IVB/SNB, where 70°C is still safe?
|
# ? Nov 25, 2012 16:53 |
|
Intel matches a slightly lower voltage with a slightly higher recommended max operating temperature. AMD specs the Phenom II 45nm up to 1.475V, but sets the top temperature to 62 C.
|
# ? Nov 25, 2012 16:57 |
|
What is kind of difference in temperature between cores is to be expected on an i5-3570k? Under 100% load, I'm seeing a 5-7c difference between the coolest and warmest core at stock clocking with water cooling. Should I be considering reapplying the TIM or is this about par for the course?
navier-stoked fucked around with this message at 05:20 on Nov 28, 2012 |
# ? Nov 28, 2012 04:45 |
That's totally normal for an IVB. Cores 1 and 2 usually run around 5-10 degrees hotter than 0 and 3 under load.
|
|
# ? Nov 28, 2012 05:00 |
|
cc posted:That's totally normal for an IVB. Cores 1 and 2 usually run around 5-10 degrees hotter than 0 and 3 under load. The difference that I'm seeing is between cores 0 and 3, curiously enough. It does seem like the temperatures are evening out a bit. Perhaps this is due to the TIM curing.
|
# ? Nov 28, 2012 06:35 |
|
drat, got my i5 3570k so I can be in the cool kids club. Computer locked up after <30 minutes of prime95 at 4.5GHz with 1.35V (I was getting game crashes at lower voltages so I raised it near max to see if it would ever be stable). Anything else I can do to stabilize it at 4.5 GHz+?
|
# ? Nov 29, 2012 02:47 |
|
Probably the only thing you could do is bring down the temperature through chip modding and crazy cooling, as heat is secondary for stability. Especially when you're dealing with a borderline-crazy overvolt like that.
|
# ? Nov 29, 2012 05:40 |
|
I have a Radeon HD 6950, if I use MSI Afterburner to OC it as well as set the voltages do I still need to turn the power control up to +20%, or does setting the voltages in Afterburner take care of that? On that note do I even need to have the graphics overdrive option ticked in Catalyst Control Center if I am using Afterburner?
|
# ? Dec 1, 2012 20:33 |
|
Destro posted:I have a Radeon HD 6950, if I use MSI Afterburner to OC it as well as set the voltages do I still need to turn the power control up to +20%, or does setting the voltages in Afterburner take care of that? On that note do I even need to have the graphics overdrive option ticked in Catalyst Control Center if I am using Afterburner? If you want to overclock beyond overdrive limits you'll probably need to follow the instructions on this page.
|
# ? Dec 2, 2012 00:55 |
|
grumperfish posted:You might need the overdrive option enabled in CCC to overclock (I don't know if this is accurate though as I use the non-CCC drivers). You have to turn power control up to +20%, especially with overvoltage, as the card will throttle once it hits TDP regardless of cooling. What do you mean by throttle? I checked the OP and it is mentioned, but not defined.
|
# ? Dec 2, 2012 02:15 |
|
First: I'm buying a i7-3770K because I write a lot of code that scales linearly with the number of cores, and i7 has 8 cores (4 virtual). However, has anyone tested how well floating point calculations scale with hyperthreading cores? With physical cores, I get a linear increase, but I'm not sure about HT "cores". Second: Can you please add this image to the first post to get people pumped
|
# ? Dec 2, 2012 04:32 |
|
Hyperthreading schedules instructions into otherwise-idle execution units of the processor. Ivy Bridge (like Sandy and Nehalem before it) has six execution ports, each with access to different execution hardware (ALU/FPU logic, bit-shift, load-store, etc.). Most of the non-memory work gets done on ports 0, 1, and 5. So if your first thread hits Port 0 and 1 at the same time, a second thread can get an instruction in on Port 5, or something can run on 0/1/5 while load/store is going on 2/3/4. In the real world, the highest performance boost you can expect is around 30% over HT disabled, with 10-20% more common. Low-threaded workloads can see a (very) minor regression because of threading overhead. Nerdwords follow: Lower level, a theoretical all-FP workload on Ivy can schedule up to three FP micro-ops at once: Port 0 can do any FP op, Port 1 can do FP add, and Port 5 can do FP move/FP logic ops. Since these ports are for decoded x86 instructions, they're not "one port per thread" or such; a 30% performance boost would mean "for the some clock ticks, a port which was otherwise empty gets filled with an op from a second thread." If you can put off your purchase, you may want to wait for Haswell, which boosts FP hardware majorly. Ports 0 and 1 get fused FMA units (FP multiply-add), fully doubling Ivy Bridge's peak FP throughput isoclock. Additionally, you'll get the option for AVX2 instructions, which double again FP throughput over AVX on Ivy, and you'll also get the TSX extensions, which can take some effort out of reducing the multithreading overhead of your software.
|
# ? Dec 2, 2012 06:02 |
|
Factory Factory posted:Hyperthreading schedules instructions into otherwise-idle execution units of the processor. Ivy Bridge (like Sandy and Nehalem before it) has six execution ports, each with access to different execution hardware (ALU/FPU logic, bit-shift, load-store, etc.). Most of the non-memory work gets done on ports 0, 1, and 5. So if your first thread hits Port 0 and 1 at the same time, a second thread can get an instruction in on Port 5, or something can run on 0/1/5 while load/store is going on 2/3/4. Thanks for the awesome response. What do you do, by the way? You have some pretty in-depth knowledge of this.. The actual CPU-intensive step in my code is solving a 4000x4000 generalized eigenvalue problem, which means inverting a matrix, multiplying by another, and then solving for the eigenvalues (all of them, but not the eigenvectors). I'm using the Intel MKL, which is the fastest compiler I've found so far. I have hundreds of these matrices to solve at each iteration, and they are independent, and so they can be farmed off to separate CPUs very efficiently. I guess I really won't know whether this can take advantage of HT or not until I try. Is there some way to step through compiled code, one instruction at a time, and see which ports are in use? Alternatively, how about a HT-emulator? I'm a physicist, not an electrical engineer, so I'm pretty much a noob at this.
|
# ? Dec 2, 2012 06:20 |
|
I'm a law student who just nerds out about this stuff, and I'm afraid my low-level nerdery and high-level nerdery don't converge quite well enough to answer your additional questions with much specificity. Everything I half-remembered from AnandTech's fantastic distillations of CPU architectures I re-referenced to the article itself. I am pretty sure, though, that the execution ports are completely abstracted away from the programmer. They're plopped behind an x86 instruction decoder (which converts x86 instructions to micro-ops) and an out-of-order scheduler. Even if you could stop on every compiled instruction, the CPU itself is still breaking it up into multiple smaller ops and running those ops mixed with ops from other x86 instructions. HT just means getting those x86 instructions from two threads at once instead of one. As for an HT emulator, I don't think such a thing would exist outside of Intel itself. I'd guess that if you wanted to test, you'd want to test on an Intel CPU with HT and a BIOS option to disable HT, like any Core i3 or i7 (or mobile i5). You must know SOMEBODY with hardware like that, or you might consider prepping a benchmark you could run at a Best Buy or Apple Store with a sympathetic staffer from the business/education relations team. Even a hyperthreaded Atom might give some idea of the gains, if any; it's only a two-port design, but it splits FP across the two ports similar to how Ivy/Sandy/Nehalem do it.
|
# ? Dec 2, 2012 07:53 |
|
Destro posted:What do you mean by throttle? I checked the OP and it is mentioned, but not defined.
|
# ? Dec 2, 2012 20:52 |
|
Factory Factory posted:IVY BRIDGE!!! I would be very curious to hear about other people's i7-3770K overclocking experiences. I'm using a Cooler Master Hyper Evo 212 CPU cooler. Running Prime95 on stock settings gives a maximum temp of about 68 C. bumping the multiplier up to 43 (from 41) gives max temps of about 80 C. This is higher than Sandy Bridge, but I've heard this is expected. I am sort-of considering using a fancier thermal paste (just using the one that came with the cooler right now), but I really doubt that will have a huge impact. Anyone else?
|
# ? Dec 7, 2012 05:47 |
|
I've got a 3570k running at 4.2Ghz at 1.2V. The hottest core reached 76C. It passed the Intel Burn Test at maximum stress which ran 10 times. Is this a decent overclock? Would you push it further?
Wooph fucked around with this message at 06:32 on Dec 9, 2012 |
# ? Dec 9, 2012 06:23 |
|
Quick specs: i5 2500k @ 4.0ghz/1.23v, CM Hyper 212+ Evo, 4 Noctua case fans, ASUS P8Z68-V PRO/GEN3, 8 gigs of Corsair RAM. I went in the BIOS and changed the multiplier and CPU voltage. I left everything else stock. I let Prime95 run overnight and my temps never went above 55c. Is it okay to leave everything else alone if I'm seemingly stable? I used that voltage because it's what a friend used for his. I know every chip is different, but I needed a reference point. I originally tried the auto-OC function through the motherboard and it put me at 4.2ghz @ 1.36v. Yikes. epic Kingdom Hearts LP fucked around with this message at 00:22 on Dec 11, 2012 |
# ? Dec 11, 2012 00:19 |
|
You are well beneath any sort of thermal or electrical walls. If you're stable at 1.24v then that isn't an issue even if you could do 1.235v or lower. Maybe you'll want to tweak around and fine-tune someday (especially if you decide to push for more clockspeed) but there's no problem there.
|
# ? Dec 11, 2012 00:28 |
|
What types of fine tuning would you recommend, and would I even notice a difference going to 4.5? A lot of the guides I've read are vague at best and I'm still new to this.
|
# ? Dec 11, 2012 00:41 |
|
I have a very similar rig and I run at 4.7GHz stable with a 0.065v offset basically all the time. There is no reason whatsoever to worry about hitting 1.36v.
|
# ? Dec 11, 2012 00:51 |
|
It's good to know that I have a lot of head room. I'm at 1.23v now and it's running great. If I feel the need for more juice later, it looks like it will be simple.
|
# ? Dec 11, 2012 00:55 |
|
|
# ? May 15, 2024 16:13 |
|
The weirdest issue I had from May until this week with the motherboard is that whenever I shut the PC down, my next first boot would ALWAYS freeze on windows loading (happened with both 7 and 8, on 2 separate SSD boot drives). I figured out through trial and error that toggling "Enable PLL Overvoltage" on was the key to getting a successful start-up, but I had to literally toggle it on after a hard reset, keeping it "on" all the time wasn't sufficient. I ran the "Extreme" power phase setting for a long time because it didn't affect heat or stability, but I recently switched to "Optimized" and for the first time since I've built it I can finally power up the first try successfully. I think this motherboard (and other Asus z68 models) still has a problem with resuming from a sleep state, which is worthy of noting.
|
# ? Dec 11, 2012 01:06 |