Overclocking a workplace PC?

[Dave3000]

I highly suggest not overclocking a work PC, whether you use one from home or at your workplace. I don't see why you would need to overclock a work PC anyways. Does the software you use at work take advantage to as many cores as you have available? If so, I suggest going with a i7-7820x or a i7-7900x. If your employer paid for that i7-7700k system as your workplace PC then I guess you are stuck with it unless you can convince him to exchange it for a Socket X299 motherboard with i7-7820x or 7900x or at the very least a 7800x. Those CPUs are better suited for heavily threaded work than the 7700k. Also the i7-7700k has a 4.4GHz stock all-core turbo, so nothing much would be gained by overclocking it anyways and definitely not worth it at the cost of instability and errors, even if it was used for just gaming. Going from a 6700k to a 7700k is like only a 5-10% performance boost, so most likely it's something you won't notice.

 

[scannall]

Also the i7-7700k has a 4.4GHz stock all-core turbo, so nothing much would be gained by overclocking it anyways and definitely not worth it at the cost of instability and errors, even if it was used for just gaming. Going from a 6700k to a 7700k is like only a 5-10% performance boost, so most likely it's something you won't notice.

If one of my employees requested an upgrade from a 6700k to a 7700k, the request would end up in the circular file. ;-)

 

[RichUK]

We do have QA before release, but this particular application's UI can't be automated so there's a limit to the amount of regression testing we can fit in. Piloting of monthly updates doesn't make sense for our customer base.

I was only messing.

 

[DaveSimmons]

I was only messing.

I didn't take it personally, but if we did deploy without QA we'd deserve as much scorn as work PC overclockers

 

[chubbyfatazn]

<snip>

You know, I remembered a similar thread being made a few years back asking essentially the same question as OP - turns out it was yours. Only remembered because of this gem (emphasis mine):

omfg NO NO NO.

Overclocking DOES NOT affect calculations.

An unstable overclock can crash a pc. The crash can cause all the issues a crash can, but when not crashing, there is no way an overclock can affect your data.

i cant believe people would write something this stupid here on AT.


your pc is doing million of calculations, always. if overclocking had a chance of producing errors, then STATISTICALLY every overclocked pc would produce a number of errors per hour. Nearly every machine owned by every person registered to this site is overclocked, NONE of them throw out errors. Nobody here has ever posted a thread "my overclock accidentally ordered ten thousand alpaca jackets from kashmir".

Any error, ANY error would make your pc useless. It's not your calculator that would fail, it would be your registry, or cache, or flie allocation table, or any number of operations your pc is doing and you are obviously ignorant of, which would fail. Digital systems do not allow for errors.
(except, funny enough, when they do. CPU-based routines do not)

A STABLE overclock is exactly the same as a stable non-overclock. There is NO factual difference between an operation performed at 4Ghz or at 3Ghz.

You do not overclock a machine attached to medical equipment because THE OVERCLOCK MIGHT NOT BE STABLE, not because the OC might produce errors.

Errors are not something you can see. You will never, ever have an error add a zero in your tax refund, because computing does not work like that. If anything, you are going to get a BSOD, and again, thats because it is not stable.

Larry, you are talking about undervolting, not overclocking. Yes there are some procedures that can cause instability, like changing the bus, overvolting, or even extreme overclocking, none of which are "a stable overclock".

At this point, i strongly suggest you return your PC to stock clock and settings IMMEDIATELY and never, ever alter another setting because obviously you have zero understanding of computing and you could hurt somebody.

Also, relevant story from Rubycon:

The problem with these popular tests is they do NOT exercise ALL cpu features. A system that seemed perfectly stable was producing errant output when encoding videos using SSE4 instructions! Whoops!

I'm not a programmer but until there is something that can do this the results from Prime, etc. just cannot be trusted. It's not how close either.

I had a chip doing 4.4GHz and erring out in a few min. Dropped to 4.2 ran OCCT all night. Encoding engines crashed once the queue started!

Dropped to 4.0GHz, same thing although the batch job did get a lot further. 3.8GHz worked however the output was checked and artifacts were found!

Finally at 3.6GHz it was completely error free. Remember this system worked perfectly with everything else at 4.2GHz! QX9650. Two QX9650's as well as QX9770's tried - the 3.2GHz 9770's were actually a little WORSE (well one of them was producing artifacts at a paltry 3.3GHz with default core).

This could be why Penryns never got above 3.2GHz.

 

[Thunder 57]

That's why you perform testing and piloting before mass deployment.

Do you think Intel shipped out a bunch of PIII 1.13GHz CPU's without testing? You can test and test again but you will never know for sure. And if someone found out that a rare error might occur because of some overclocked PC (or any other generally accepted bad idea), heads will likely roll.

EDIT

It doesn't even have to be overclocking. Say you custom build a few PC's for some department because it's cheaper that way. Then, any little problem they may have, guess who they're coming to? They'll come to you if it's an OEM one to, but at least then it's their problem to fix.

Say there is a Windows update or driver problem that occurs, and you don't have the answer right away. You might get stuck dealing with subpar MS or mobo vendor support, etc. They get pissy and bitch to someone higher up, they come to you and ask why the hell they are using custom built PC's by Joe the IT guy.

Dell/HP/etc spend a lot of money doing testing and validation to make sure their stuff works together. They are putting their reputation on it. That's why they cost more. You don't want to save some money in hardware then have a problem with a custom build that takes some time to resolve, all the while losing much more money than was saved due to lack of production.

/End Rant

 

[StefanR5R]

All this talk about mission-critical workstations which supposedly need to be super-reliable. And yet these are usually cheap mainstream PCs (maybe bundled with costly support fees though), or laptops even, instead of traditional workstation hardware.

I am getting though that this song and dance is not about actual reliability, but about keeping the hardware vendors (as much as possible) responsible for hardware reliability. Software vendors in contrast succeed to avert this kind of responsibility mostly.

Edit:
And commenters continue to leave aside that organizations which deploy 10 PCs may have different requirements vs. risks than ones which deploy 1,000s of PCs.

 

[Thunder 57]

All this talk about mission-critical workstations which supposedly need to be super-reliable. And yet these are usually cheap mainstream PCs (maybe bundled with costly support fees though), or laptops even, instead of traditional workstation hardware.

I am getting though that this song and dance is not about actual reliability, but about keeping the hardware vendors (as much as possible) responsible for hardware reliability. Software vendors in contrast succeed to avert this kind of responsibility mostly.

Edit:
And commenters continue to leave aside that organizations which deploy 10 PCs may have different requirements vs. risks than ones which deploy 1,000s of PCs.

I think your definition of mission-critical and mine may be different. When I hear that term I'm thinking ECC, redundant PSU, etc.

For everyday work it is still critical that a user has a computer, but it can fail. When it does the best thing is to have an identical model on hand that you can replace it with. Once you start doing custom builds, that becomes a lot more difficult to manage. This is forgetting the whole overlocking part which this thread is/was about.

 

[IndyColtsFan]

All this talk about mission-critical workstations which supposedly need to be super-reliable. And yet these are usually cheap mainstream PCs (maybe bundled with costly support fees though), or laptops even, instead of traditional workstation hardware.

I am getting though that this song and dance is not about actual reliability, but about keeping the hardware vendors (as much as possible) responsible for hardware reliability. Software vendors in contrast succeed to avert this kind of responsibility mostly.

Edit:
And commenters continue to leave aside that organizations which deploy 10 PCs may have different requirements vs. risks than ones which deploy 1,000s of PCs.

That's because again, you're not getting it. I'll also address your previous reply to me:

1. Others have defined production in this thread several times, so I won't beat that horse to death. You can throw out as many niche situations or examples as you want, but we're talking about a guy with his PC and apparently custom building a replacement.
2. You keep talking about the workloads he runs. And how do YOU know the workloads he runs exactly? Here is the FACT: He upgraded from a 6700 to a 7700. There is VERY little performance to be gained by doing so. Chances are that his "sluggish" performance was due not to the CPU, but other factors but if, in fact, his slow speeds ARE because of his workloads, how would a "slow" 6700 be radically improved with a 7700 exactly?
3. Yes, I know that certain cloud providers use desktop grade hard drives but again, I do not think you fully understand. You're correct that they plan and compensate for the higher failure rates, but what you're missing is that IT operations are their CORE business. It is much different for AWS to maintain a spare inventory of hundreds of desktop drives for quick replacements than it is for IT departments where IT isn't the core competency (insurance, manufacturing, etc)
4. The largest corporations also tend to buy pre-built servers from the likes of Dell and HP. Why? Because it is typically cheaper, spare parts are available everywhere, and you can get super fast service. Who wants to pay people to spend their days building servers from the ground up? Very, very few. You may have some situations where companies use white box builds for VMWare labs or test environments, but those aren't production instances either.
5. Of all companies, small businesses are the least equipped to build servers and systems from scratch. Oh, trust me, I DO know it happens but it is penny wise and pound foolish IMO. Which business is the least likely to be able to afford a user being down for 1 day? A small business or a Fortune 500 company?
6. You've mentioned hardware reliability and that's part of it, but you're overlooking the huge support advantages of using systems from a large vendor like HP or Dell. Problems with your system? There are official support forums and likely thousands or millions of folks using the same system as you who can assist on third-party forums (like AT). And in the realm of servers, you can get parts replaced in HOURS - even system boards!
7. Yes, there are often monetary advantages of sticking with one vendor. There is nothing wrong with leveraging that.
8. You're overlooking the support headaches involved when you have users who are allowed to run amok and build/overclock/maintain their own PCs. Anyone who has worked in IT for any length of time always has a niche group of self-proclaimed "computer experts" in their user base who work in other departments. You know the type - they say stupid stuff like this:

"EXPERT" #1: "What?!?!?! You pay how much for our internet connectivity? I pay $50/month at Comcast and get 100 MB speeds! I'm going to put a bug in the CEO's ear that we should get a cable modem and use Comcast to save money!"
"EXPERT" #2: "We pay HOW MUCH for these laptops? I bought an Acer at Best Buy for $350 and I can game on it! We should call Best Buy and use them as our supplier and use Acer PCs!"
"EXPERT" #3: <pulls the IT guy aside and speaks in a low, barely audible voice> "I took a computer class on Saturday mornings at the local community college, so I'm an expert and am "one of you." If you guys need any support help, just let me know, and maybe you can let me "test" some non-standard software, if you know what I mean." <winks at IT guy>

The typically clueless business PC user isn't your worst enemy in IT - it is the person who, for whatever reason, thinks he/she is a computer expert and then when they do stupid stuff and it breaks things, they literally trip over themselves running to IT to fix their problems. "What? Why, I don't have any idea how that extra RAM module got installed in my machine! You guys must have done it but all this crashing lately seems to me to be bad memory."

I digress, but I can't help but tell about a true story I encountered many years ago. Even some other IT folks are idiots who know enough to be danger. A little over a decade ago, I worked in an IT department at a large company. I had a friend there who was a programmer but constantly bragged about how he was an "expert" on computers and programming. He literally told anyone who would listen how he was the "best" and "only true" IT person because he has a degree in Computer Technology (no, I'm not joking). Anyone, a couple of memorable examples from this dumbass:

1. He needed Office installed on his test development box. The help desk was way too busy to do it and they (the help desk) called me. I said "Give him the Office CD and tell him to install it himself - he is in IT and is a self-proclaimed expert, plus anyone can install Office." They called me back like 20 minutes later: "He wants to know if we have instructions, as he says he doesn't know how to install office." WTF
2. I tend to think he thought very highly of himself partially because this was his first IT job (IIRC) and he had never really worked with anyone with a clue. He was sitting in my office one day bragging about a fixing a long-standing issue one of his programs had. Him: "Yeah, I fixed that problem - what I did was make the value a variable instead of hardwiring it in the code." He was very proud of himself and acted like he just solved world hunger. But I ask you - read his statement again - doesn't anyone with even a high school level of programming know stuff like that? I was embarrassed for him.

 

[StefanR5R]

You can throw out as many niche situations or examples as you want,

I felt a need to do so because others stated ill-defined generalities as if they were true universally. However, I think at this point in the discussion it has become obvious that such considerations don't just follow simple universal wisdoms, but are results of cost-benefit evaluations, or at least should be.

E.g., regarding the particular point #2: The only argument I wanted to make in this regard is that a certain hardware reliability feature* should IMO not be painted as (objectively?) not required, but rather (without deeper analysis: subjectively) too costly relative to the risks involved in his workload (which are unknown to me and you, unless you know more about his jobs and contractual relationships).

*) ECC RAM

 

[Comdrpopnfresh]

I am extremely pleased with the responses, and it is one of the most fruitful threads I have started in these forums that did not devolve into flaming or unintelligent written assault.

No one looks over my shoulders for what I do with the workstation. The only risk, which I will not approach, is getting hanging tasks or BSODs that impede my ability to get things done. I personally feel that modestly gaining more performance a system is readily capable of, when the workplace settings allow, is better than having to compile and distribute executables from my IDE workspace to run on remote machines because the crummy beige-box (disclosure- it's black and grey) that came from a vendor chokes on the regular.

 

[Ratman6161]

I am extremely pleased with the responses, and it is one of the most fruitful threads I have started in these forums that did not devolve into flaming or unintelligent written assault.

No one looks over my shoulders for what I do with the workstation. The only risk, which I will not approach, is getting hanging tasks or BSODs that impede my ability to get things done. I personally feel that modestly gaining more performance a system is readily capable of, when the workplace settings allow, is better than having to compile and distribute executables from my IDE workspace to run on remote machines because the crummy beige-box (disclosure- it's black and grey) that came from a vendor chokes on the regular.

Well, you might think those are the only risks...but remember its all fun and games until you get audited. You may not worry about this if you are in a very small company. But I've been finding out over the last two years about how financial auditors now often show up with their own cadre of IT auditors. They want to see every i dotted and every t crossed as far as having every single piece of software licensed and paid for and no, your word on it will not be good enough. They also want to see every PC fully inventoried for both hardware and software. They want documented proof that you have a documented configuration standard and that your PC's and servers all follow it. We even got written up because their network scans detected printers with out of date firmware (no joke, this is real). They want to see docuemented evidence that every system has antivirus etc...and all that is before they even get to password policies, network protocols etd.

And we are not Fortune 500 either. We are talking about a company of only about 140 employees. You essentially can't pass an audit anymore unless all of your systems are centrally managed. There just isnt any other way to get there. Also, we don't fix PC's. We fix software configuration issues but the moment we determine we have a hardware problem its a warranty issue. We have extra PC's and if there is anything seriously wrong with yours we yank it, give you a spare and call our vendor with whom we have next day on site support. They come and fix it and you are up and running in minutes.

To make all this work, we need all the PC's to be as identical as we can make them. We tend to go upscale on them in that they are all i7 7700's (no "K" chip is getting into my shop) with 16 GB of RAM. There are some allowances made for our developers in that we start them out with the standard configuration and temporarily give them permissions to install so they can install and configure their tools, but that's about it.

Don't get me wrong. I love building my own systems...at home. I also love overclocking...at home on the stuff I paid for and support myself. But the last time I built PC's for work was in 1999. I would never get away with it now even if I wanted to and now that I'm in charge of IT, I'd never let anyone else do it either.

 

[jmelgaard]

...because he has a degree in Computer Technology...

Are you sure the HR department checked thoroughly where he had his degree from? >.<... sounds like it came from a vending machine...

Our IT department actually lets us install extra RAM and SSD's our self (if we are ok with it) under the general assumption that "Most developers can do that"...
However the hardware is still provided by them (and in turn from our supplier, with some exceptions, currently Lenovo)...

 

[Smoblikat]

Why not? Clocks can be limited by power delivery, cooling capacity, and reliability. On multicore CPUs subjected to multicore workload, there is a very well known overclocking headroom within reliability limits, if power delivery and cooling capacity are provided accordingly.

Not exactly. The only well known headroom comes from the chips built in turbo frequencies, and only under specified conditions too. Anything other than stock and youre not guarenteed anything but a headache when it comes to production environments. If you value your data being uncorrupted and you value the overall integrity of your system, never OC an important computer. That being said I always try to run setfsb on my personal work computers, but only the ones that arent mission critical. and yes, I do have a backup.

 

[urvile]

That's because again, you're not getting it. I'll also address your previous reply to me:

6. You've mentioned hardware reliability and that's part of it, but you're overlooking the huge support advantages of using systems from a large vendor like HP or Dell. Problems with your system? There are official support forums and likely thousands or millions of folks using the same system as you who can assist on third-party forums (like AT). And in the realm of servers, you can get parts replaced in HOURS - even system boards!

I remember years ago I was working for a small business that sourced all of their laptops from Dell with top support coverage. One day I spilled (a lot) of coke on the keyboard of my latop and turned it into a big sticky mess. So I contacted Dell and asked them if I could purchase a new keyboard. They got back to me and said these things happen we will send a tech to your office to replace it under warranty and they did.

That was some impressive service. Dell's warranty even covers clumsy idiots apparently.

 

[pjmssn]

We only purchase Dell computers (and Apple) at the university. The IT department requests the computers to be shipped to them directly where the BIOS are locked and the computers set up. The faculty members are allowed to have administrator accounts and we can install any software freely, but the students computers have to go to IT for software installation.
It is not convenient and wastes a lot of time, but the system is very secure and the required support minimum since (theoretically) no malware can be installed.
So, no OC and standard supported computers only, mostly precisions and optiplex. All the hardware support is handled by the IT department, including upgrades or warranty work.

 

[Smoblikat]

I remember years ago I was working for a small business that sourced all of their laptops from Dell with top support coverage. One day I spilled (a lot) of coke on the keyboard of my latop and turned it into a big sticky mess. So I contacted Dell and asked them if I could purchase a new keyboard. They got back to me and said these things happen we will send a tech to your office to replace it under warranty and they did.

That was some impressive service. Dell's warranty even covers clumsy idiots apparently.

Thats nothing, I used to work for internal IT support for Dell. I once had a guy spill an entire bottle of bourbon on his computer (the bottle was in his laptop bag and it broke open). I had to replace every single component in the computer, except the bare chassis (which needed to be seriously cleaned with soap and water). I mentioned to my boss that it would definitely be faster, easier, and cheaper to just give him a new computer, but Dell has an internal warranty of 3 years, and its expected that that warranty is honored no matter what.