[SOLVED] Voltage and Clock Speeds

Mar 4, 2019
25
0
30
0
Is there some sort of general "formula" for finding the lowest voltage required for a processor to run a certain clock speed stably? I don't want to run the voltage higher than is necessary for a specific clock speed, so I don't use extra power or generate excess heat, or probably fry the whole thing.
 

Darkbreeze

Titan
Moderator
Yes, trial and error. Generally speaking, it MUST be an incremental process for exactly the reasons you've outlined. Small increase in clock frequency, small increase in voltage, test. Stable? Then another small increase in clock frequency and test. Not stable? Then small increase in voltage and re-test. There is NO formula for simply jumping to a specific clock speed, slapping a predetermined voltage on it and knowing it will be stable. You have to do the work.

Quick and dirty overview of overclocking/stability validation procedure.

Set CPU multiplier and voltage at desired settings in BIOS. Do not use presets or automatic utilities. These will overcompensate on core and other voltages. It is much better to configure most core settings manually, and leave anything left over on auto until a later point in time if wish to come back and tweak settings such as cache (Uncore) frequency, System agent voltage, VCCIO (Internal memory controller) and memory speeds or timings (RAM) AFTER the CPU overclock is fully stable.

Save bios settings (As a new BIOS profile if your bios supports multiple profiles) and exit bios.

Boot into the Windows desktop environment. Download and install Prime95 version 26.6.

Download and install either HWinfo or CoreTemp.

Open HWinfo and run "Sensors only" or open CoreTemp.

Run Prime95 (ONLY version 26.6) and choose the "Small FFT test option". Run this for 15 minutes while monitoring your core/package temperatures to verify that you do not exceed the thermal specifications of your CPU.

(This should be considered to be 80°C for most generations of Intel processor and for current Ryzen CPUs. For older AMD FX and Phenom series, you should use a thermal monitor that has options for "Distance to TJmax" and you want to NOT see distance to TJmax drop below 10°C distance to TJmax. Anything that is MORE than 10°C distance to TJmax is within the allowed thermal envelope.)

If your CPU passes the thermal compliance test, move on to stability.

Download and install Realbench. Run Realbench and choose the Stress test option. Choose a value from the available memory (RAM) options that is equal to approximately half of your installed memory capacity. If you have 16GB, choose 8GB. If you have 8GB, choose 4GB, etc. Click start and allow the stability test to run for 8 hours. Do not plan to use the system for ANYTHING else while it is running. It will run realistic AVX and handbrake workloads and if it passes 8 hours of testing it is probably about as stable as you can reasonably expect.

If you wish to check stability further you can run 12-24 hours of Prime95 Blend mode or Small FFT.

You do not need to simultaneously run HWinfo or CoreTemp while running Realbench as you should have already performed the thermal compliance test PLUS Realbench will show current CPU temperatures while it is running.

If you run the additional stability test using Prime95 Blend/Small FFT modes for 12-24 hours, you will WANT to also run HWinfo alongside it. Monitor HWinfo periodically to verify that no cores/threads are showing less than 100% usage. If it is, then that worker has errored out and the test should be stopped.

If you find there are errors on ANY of the stability tests including Realbench or Prime95, or any other stress testing utility, you need to make a change in the bios. This could be either dropping the multiplier to a lower factor or increasing the voltage while leaving the multiplier the same. If you change voltage or multiplier at ANY time, you need to start over again at the beginning and verify thermal compliance again.

A more in depth but general guide that is still intended for beginners or those who have had a small amount of experience overclocking can be found here:


 
Mar 4, 2019
25
0
30
0
Yes, trial and error. Generally speaking, it MUST be an incremental process for exactly the reasons you've outlined. Small increase in clock frequency, small increase in voltage, test. Stable? Then another small increase in clock frequency and test. Not stable? Then small increase in voltage and re-test. There is NO formula for simply jumping to a specific clock speed, slapping a predetermined voltage on it and knowing it will be stable. You have to do the work.

Quick and dirty overview of overclocking/stability validation procedure.

Set CPU multiplier and voltage at desired settings in BIOS. Do not use presets or automatic utilities. These will overcompensate on core and other voltages. It is much better to configure most core settings manually, and leave anything left over on auto until a later point in time if wish to come back and tweak settings such as cache (Uncore) frequency, System agent voltage, VCCIO (Internal memory controller) and memory speeds or timings (RAM) AFTER the CPU overclock is fully stable.

Save bios settings (As a new BIOS profile if your bios supports multiple profiles) and exit bios.

Boot into the Windows desktop environment. Download and install Prime95 version 26.6.

Download and install either HWinfo or CoreTemp.

Open HWinfo and run "Sensors only" or open CoreTemp.

Run Prime95 (ONLY version 26.6) and choose the "Small FFT test option". Run this for 15 minutes while monitoring your core/package temperatures to verify that you do not exceed the thermal specifications of your CPU.

(This should be considered to be 80°C for most generations of Intel processor and for current Ryzen CPUs. For older AMD FX and Phenom series, you should use a thermal monitor that has options for "Distance to TJmax" and you want to NOT see distance to TJmax drop below 10°C distance to TJmax. Anything that is MORE than 10°C distance to TJmax is within the allowed thermal envelope.)

If your CPU passes the thermal compliance test, move on to stability.

Download and install Realbench. Run Realbench and choose the Stress test option. Choose a value from the available memory (RAM) options that is equal to approximately half of your installed memory capacity. If you have 16GB, choose 8GB. If you have 8GB, choose 4GB, etc. Click start and allow the stability test to run for 8 hours. Do not plan to use the system for ANYTHING else while it is running. It will run realistic AVX and handbrake workloads and if it passes 8 hours of testing it is probably about as stable as you can reasonably expect.

If you wish to check stability further you can run 12-24 hours of Prime95 Blend mode or Small FFT.

You do not need to simultaneously run HWinfo or CoreTemp while running Realbench as you should have already performed the thermal compliance test PLUS Realbench will show current CPU temperatures while it is running.

If you run the additional stability test using Prime95 Blend/Small FFT modes for 12-24 hours, you will WANT to also run HWinfo alongside it. Monitor HWinfo periodically to verify that no cores/threads are showing less than 100% usage. If it is, then that worker has errored out and the test should be stopped.

If you find there are errors on ANY of the stability tests including Realbench or Prime95, or any other stress testing utility, you need to make a change in the bios. This could be either dropping the multiplier to a lower factor or increasing the voltage while leaving the multiplier the same. If you change voltage or multiplier at ANY time, you need to start over again at the beginning and verify thermal compliance again.

A more in depth but general guide that is still intended for beginners or those who have had a small amount of experience overclocking can be found here:


so that means disabling turbo boost, right?
 

Darkbreeze

Titan
Moderator
Depends on the platform. For modern Intel platforms, you need to leave the boost features enabled because that's actually how you get the per core overclock to work on most of them. For older AMD platforms, you'd disable it and do everything via CPU multiplier. For AMD Ryzen platforms, I'm not sure which option is preferable as I haven't yet done any overclocking myself on Ryzen. If it is a Ryzen platform though, there is certainly TONS of information available on that.

What are your hardware specifications for all your core hardware like motherboard, cpu, power supply, memory, etc.?
 
Mar 4, 2019
25
0
30
0
Depends on the platform. For modern Intel platforms, you need to leave the boost features enabled because that's actually how you get the per core overclock to work on most of them. For older AMD platforms, you'd disable it and do everything via CPU multiplier. For AMD Ryzen platforms, I'm not sure which option is preferable as I haven't yet done any overclocking myself on Ryzen. If it is a Ryzen platform though, there is certainly TONS of information available on that.

What are your hardware specifications for all your core hardware like motherboard, cpu, power supply, memory, etc.?
https://www.userbenchmark.com/UserRun/15183456
the specs are there. also, i might have gotten some insane performance from my cpu, but its not even close to stable. i was lucky to have it even run the benchmark successfully without crashing the pc.

also, how do you figure out if you increase the voltage too much. unless if the only way is by frying the cpu.
 

Darkbreeze

Titan
Moderator
The only way you'd "fry your CPU" is if you try jumping to a specific configuration without doing the work.

You "figure out" if you've increased the voltage too much by thermal testing. If you exceed 80°C at any point, then the voltage is too high or there is a problem with the cooling such as no enough case airflow, not capable enough CPU cooler or poorly mounted CPU cooler. The steps are clearly outlined in the post I made earlier AND in my basic overclocking tutorial I linked to.
 
Mar 4, 2019
25
0
30
0
The only way you'd "fry your CPU" is if you try jumping to a specific configuration without doing the work.

You "figure out" if you've increased the voltage too much by thermal testing. If you exceed 80°C at any point, then the voltage is too high or there is a problem with the cooling such as no enough case airflow, not capable enough CPU cooler or poorly mounted CPU cooler. The steps are clearly outlined in the post I made earlier AND in my basic overclocking tutorial I linked to.
Ok. Thanks for the help! I got some idea from some other sources that I could destroy my CPU just by a big enough voltage increase, not by dangerous amounts of heat. But I guess by know I should have fully realized that increased voltage and heat are directly connected...

Sorry, I forgot to mention it also has a Cooler Master 600W Power Supply.
 

Darkbreeze

Titan
Moderator
What is the EXACT model of your power supply?

Yes, there IS generally a "death voltage", and you can certainly kill a CPU by using too much voltage but you'll never reach that point if you follow the proper steps and incrementally verify at each stage of the process. You'll find the plateau where the CPU either becomes unstable or you can't remain within the thermal specifications, long before you reach the death voltage that would quickly cause electrical damage to the VRMs or CPU.
 
Mar 4, 2019
25
0
30
0
What is the EXACT model of your power supply?

Yes, there IS generally a "death voltage", and you can certainly kill a CPU by using too much voltage but you'll never reach that point if you follow the proper steps and incrementally verify at each stage of the process. You'll find the plateau where the CPU either becomes unstable or you can't remain within the thermal specifications, long before you reach the death voltage that would quickly cause electrical damage to the VRMs or CPU.
My PC is a PowerSpec G351. All I got from the specs page is that the power supply is a "600 80 plus". Is that enough? I don't know if the only version of that is the bronze one, or if there are different versions of it.
 

Darkbreeze

Titan
Moderator
ON the power supply itself there is a specifications label. Has to be, it's the law. The model of the power supply will be listed on the specifications label. It will be on one of the four sides. Sometimes, rarely, you might have to remove the four screws that hold it to the back of the case and turn the power supply to read the label.
 
Mar 4, 2019
25
0
30
0
ON the power supply itself there is a specifications label. Has to be, it's the law. The model of the power supply will be listed on the specifications label. It will be on one of the four sides. Sometimes, rarely, you might have to remove the four screws that hold it to the back of the case and turn the power supply to read the label.
Lemme guess. It's on a sticker.

On the side facing the glass panel of my rig, it says "Cooler Master MasterWatt Lite 600". According to the specs page, it has a 80 PLUS Standard certification
 

Darkbreeze

Titan
Moderator
Let me tell you something about that unit, because that unit is SPECIFICALLY included in my normal power supply recommendation copy pasta, and it's not good. There is a link to the reviews of it included. I'd HIGHLY recommend that you replace it with a high quality unit before you do any overclocking, or even if you never overclock at all. It's an exceptionally poor quality power supply.

Click the spoiler below to read the recommendations and the recommendation AGAINST that unit, or most Cooler master power supplies for that matter.

Let's start with the biggest misconception out there, which is that if a unit has high watts it will be ok or is good. No. Just, no.

There are plenty of 750-1000w units out there that I wouldn't trust to power a light bulb and might in fact be more dangerous due to their supposedly high capacity due to poor or non-existent protections inside the unit.

If the platform isn't good to begin with, how many watts or amps it says it can support is irrelevant.

Higher 80plus certification doesn't mean anything, UNLESS it's on an already known to be high quality PSU platform. For example, a Seasonic Prime platinum unit is going to be a better product than a Seasonic Prime Gold unit, because we already know the Prime platform is very good, and platinum efficiency along with it shows there are some improvements internally to account for the higher efficiency.

In a case like that, it might be worth it. It's likely the unit will create less heat, it will probably have better performance in regard to ripple, noise and voltage regulation. It might shave a few pennies, or dollars, off the electric bill over the course of a year.

Other than that, it is not going to perform any better than the same platform with Gold efficiency. On the other hand, just because a unit has Titanium 80plus ratings doesn't mean the unit is any good at all. For example, there are Raidmax units with Titanium efficiency and I wouldn't trust one of those to power a light bulb. There are a lot of units like this out there.

If the platform isn't good to begin with, whether or not it has an 80plus certification or not is irrelevant.

Whatever you do, don't EVER buy a power supply based on whether it has RGB or lighting, or looks like it might be a quality unit. Some of the biggest hunks of junk out there look just as good as a Seasonic Prime Ultra Titanium, but I assure you, they are not. So far as I've seen there are really no excellent units out there that have RGB built in. Maybe one or two models, but rest assured you'll be be paying for the lighting, not for the quality of the power supply.

I don't know what country you reside in, and I know that sometimes it's hard to come by good units in some regions, but when possible, when it comes time to get that PSU, I'd stick to the following if you can.

Seasonic. Seasonic isn't just a brand, they are a PSU manufacturer, unlike many of the PSU brands you see they make their own power supply platforms AND a great many of the very good PSU models out there from other brands like Antec, Corsair and older XFX are made by Seasonic.

Just about anything made by Seasonic is good quality for the most part. There are really no bad Seasonic units and only a very few that are even somewhat mediocre. They do make a few less-good quality OEM style units, but mostly those are not going to be units you come across at most vendors, and they are still not bad. Also, the S12II and M12II 520 and 620w units are older, group regulated models. At one time they were among the best units you could buy. Now, they are outdated and not as good as almost any other Seasonic models. They are however still better than a LOT of newer designs by other manufacturers.

The Seasonic 520w and 620w S12II/M12II units CAN be used on newer Intel platforms, if you turn off C6/C7 in the bios, but I'd really recommend a newer platform whenever possible. Prices are usually pretty good on those though, so sometimes it's worth accepting the lack of DC-DC on the internal platform. Higher capacity versions of the High current gamer are not based on that platform, so they are fine. Those being the 750w and higher versions.

Most common currently, in order of preference, would be the Seasonic Focus series, then Focus plus, then Prime, then Prime ultra. It's worth mentioning that there are generally Gold, Platinum and Titanium versions within each, or most, of those series, but that does not necessarily mean that a Focus plus Platinum is necessarily better than a Prime Gold. It only means that it scored better in the 80plus efficiency testing, not that the platform is better.

Again, don't let yourself get tangled up in the idea that a higher 80plus rating specifically means that it is a better unit than another one with a lower rating, unless you know that it is a good platform from the start. All these Focus and Prime units are pretty good so you can somewhat focus on the 80plus rating when deciding which of them to choose.

Super Flower Super Flower is another PSU manufacturer. They also make most of the good units sold by EVGA like the G2, G3, P2 and T2 models.

Super Flower doesn't have a very broad availability for the units with their own brand name on them, and are not available in a lot of countries but for those where there is availability you want to look at the Leadex and Leadex II models. The Golden green platform is fairly decent too but is getting rather long in the tooth as a platform AND I've seen some reviews indicating a few shortcomings on units based on this platform.

Even so, it's a great deal better than a lot of other platforms out there so you could certainly do worse than a Golden green model. Units based on the Leadex and Leadex II platforms are much better though.

Corsair. The CX and CXm units are ok as a budget option, but I do not recommend pairing them with gaming cards. The newer 2017 models of CX and CXm are better than the older ones, but still not what we'd call terrific, so if it specifically says 2017 model, or it has a capacity other than an even 100, like 550w, 650w, 750w, etc., then it's likely at least better than those older ones. Aside from that, any of the TX, RMx, RMi, HX, HXi, AX or AXi units are good. Those are listed from best to worst, with the best being the AX and AXi units.

Antec. The True power classic units are made by Seasonic, and are very good, but are not modular. The High current gamer 520w and 620w, or any other PSU you see on the market that is 520w or 620w, are also made by Seasonic, based on the S12II and M12II platform for modern versions, and are pretty good units but again they are an older platform that is group regulated so if you go with a Haswell or newer Intel configuration you will want to avoid those because they do not support the C6/C7 Intel low power states.

The Antec High current gamer 750w and 850w units are very good and are not the older design, which came in 520w and 620w capacities and were good for back then but again, are an aging Seasonic platform that is not the best choice most of the time these days. Occasionally, these older units MIGHT be the best unit available and you could do worse than one of them, but a newer DC-DC platform is desirable when possible if it doesn't mean sacrificing quality elsewhere in the platform. There are however older and newer HCG models, so exact model number will likely be a factor if choosing one of these however both the older models and the newer models are good.

Antec Edge units are ok too, but reviews indicate that they have noisy fan profiles. I'd only choose this model if it is on sale or the aesthetics match up with your color scheme or design. Still a good power supply but maybe a little aggressive on the fan profile. This may have been cured on newer Edge models so reading professional tear down reviews is still the best idea.

Antec Earthwatts Gold units are very good also.

BeQuiet. BeQuiet does have a few decent models, BUT, you must be VERY selective about which of their models you put your trust in. From model to model their are huge differences in both quality and performance, even with the same series. If you cannot find a review for a BeQuiet unit on HardOCP, JonnyGuru or Tom's hardware that SPECIFICALLY says it is a very good unit, and does not have any significant issues in the "cons" category, I would avoid it. In fact, I'd probably avoid it anyhow unless there is a very great sale on one that has good reviews, because their units are generally more expensive than MUCH better units from Antec, Seasonic, EVGA and Corsair.

Super Flower. They are like Seasonic and they make power supplies for a variety of other companies, like EVGA. Super Flower units are usually pretty good. I'd stick to the Leadex, Leadex II and Golden Green models.

EVGA. They have BOTH good and not very good models.

Not very good are the W1, N1, B1, B3 (All models except the 650w model), BQ, BR, BT and G1 NEX models.

Good models are the B2, B3 650w, G2, G2L, G3, GQ, P2 and T2 models.

FSP. They used to be very mediocre, and are a PSU manufacturer like Seasonic and Super Flower, although not as well trusted based on historical performance. Currently the FSP Hydro G and Hydro X units are pretty good.

I would avoid Thermaltake and Cooler Master. They do have a few good units, but most of the models they sell are either poor or mediocre, and the ones they have that ARE good are usually way overpriced.

This is just ONE example of why I say that. Very new and modern CM unit. One of the worst scores ever seen on JonnyGuru for a well known brand name product. Doesn't look to be much better than a Raidmax unit. Sad.


And most of the models I have linked to the reviews of at the following link are at least good, with most of them being fantastic.


Certainly there ARE some good units out there that you won't see above among those I've listed, but they are few and far between, much as a hidden nugget of gold you find in a crevice among otherwise ordinary rocks and don't EVER assume a unit is good just because of the brand.

If you cannot find an IN DEPTH, REPUTABLE review on Tom's hardware, JonnyGuru, HardOCP, Hardware secrets (Old reviews by Gabe Torres), Kitguru (Only Aris reviews), TechPowerUP, SilentPC crew or a similar site that does much more than simply a review of the unboxing and basic tests that don't include reliable results for ripple, noise, voltage regulation and a complete teardown of the unit including identification of the internal platform, then the unit is a big fat question mark.

I recommend not trusting such units as companies generally always send out review samples of any unit they feel is going to get a good review, and don't send them out if they know they are going to get hammered by the reviewer. No review usually equals poor quality. Usually.

Other models that should never be trusted OR USED AT ALL, under any circumstances, include A-Top, AK Power, Apevia, Apex (Supercase/Allied), Artic, Ace, Aerocool (There might be one model worth using, but I'd still avoid them.), Aspire (Turbocase), Atadc, Atrix, Broadway com corp, Chieftech, Circle, CIT, Coolmax, Deer, Diablotek, Dynapower, Dynex, Eagletech, Enlight, Eurotech, Evo labs, EZ cool, Feedtek, Foxconn, G7, HEC/Compucase Orion, HEDY, High power, iBall, iStar computer co., Jeantec, JPac, Just PC, Kolink, LC Power, Linkworld electronics, Logisys, Macron, MSI, NmediaPC, Norwood Micro (CompUSA), Okia, Powercool, Powmax, Pulsepower, Q-tec, Raidmax, RaveRocketfish, Segotep, SFC, Sharkoon, Shuttle, Skyhawk, Spire, Startech, Storm, Sumvision, Tesla, Trust, Ultra, Wintech, Winpower, Xilence (Until I see a reputable review of a model showing different), xTreme (Cyberpower), Youngbear and Zebronics.
 

ASK THE COMMUNITY

TRENDING THREADS