The Witcher 3: Wild Hunt

[ImpulsE69]

Yes I know that which is why I mentioned previously that someone overclocking needs to find the frequency that works for their hardware. Someone with my same cpu can get 4.7 and my chip won't. The voltage required might be different too.

I am contesting the idea that "no overclock is stable ever". That's false

That's a case of you trying to be way too literal on the internet. There are always caveats to scenarios, even what you may deem stable overclocks. The only true (one would hope) stable clock is the stock clock.

Hopefully I don't have to explain why I facepalmed the other guys post. It was funny watching futurefields pretend like he had a clue though.

 

[cmdrdredd]

Again, stable is stable. If it's 3Ghz or 4.5Ghz it doesn't matter.

 

[ImpulsE69]

Again, stable is stable. If it's 3Ghz or 4.5Ghz it doesn't matter.

Until it isn't.

I ask this. How do you determine if an overclock is stable or not?

The answer is the answer.

 

[cmdrdredd]

You run all the software available that has built in checks. If it passes everything then it's stable. It's not that difficult and I don't know why you seem hell bent on proving that anyone who overclocks has an unstable machine.

The problem is that some programs are unstable to begin with which is why there are established and accepted standards for considering a system 24/7 stable. These accepted standards go way beyond any stress a game or working in photoshop or transcoding videos can do to a system.

 

[Chaotic0ne]

Anyone with an AMD GPU that's Radeon 7XXX series or newer might wanna get this performance DLL file http://www.nexusmods.com/witcher3/mods/104

I tried it earlier, and I used the 1040 pack. At first I didn't think it was getting me any gains, but the increase was quite noticeable in some spots where I saw as much as 6-8fps higher. Such as spots that were 52fps without the DLL were pushed to 60. With all the settings maxed out, including hairworks with x8 AA, I can get 45-50fps running around Novigrad. That's with a 1130/1700 OC on my R9 390. This DLL is supposed to help in CPU intensive spots. So depending on the CPU you got, you might not see any gains, but an I5 2500K @ 4.2ghz bottlenecks occasionally on Witcher 3.

 

[Merad]

Hopefully I don't have to explain why I facepalmed the other guys post. It was funny watching futurefields pretend like he had a clue though.

By all means, explain to me what happens when software "has an issue with overclock." I'll be getting my popcorn.

 

[ImpulsE69]

You run all the software available that has built in checks. If it passes everything then it's stable. It's not that difficult and I don't know why you seem hell bent on proving that anyone who overclocks has an unstable machine.

The problem is that some programs are unstable to begin with which is why there are established and accepted standards for considering a system 24/7 stable. These accepted standards go way beyond any stress a game or working in photoshop or transcoding videos can do to a system.

Yes and no. Overclocks are typically stable in a vacuum, but that isn't what you are using them for. This is why you can overclock a system higher in bios, then it hangs while going into Windows. Or why you can make it into windows, browse the internet, but you try to load a game and you crash. It's all about the software interaction with the hardware. The tests you do make it MORE likely your overclock will work in any situation, but it doesn't mean it will in all cases. It's more complicated than that. There's a study done by MS out there somewhere about the rates of errors based on OC vs no OC, and how it relates also to errors in other portions of the hardware, but I can't find it right now. Granted if I recall, it wasn't a very good study, but there some explanations there that are useful.

When you say overclocks are unstable it is because they always are. They are only as stable as the software being run on them + everything else happening on the system. Thus why we say "such and such doesn't play nice with overclocks". Like you said some programs are unstable, and that may be the root of the issue of scenarios such as these, but sometimes the fix or workaround is to not overclock - UNTIL (or if they ever do) they fix whatever is causing it to not work with the OC. This can represent a very small portion of people, based on certain criteria. It does not at all mean OMG OC's don't work run for the hills. It just means that there may be a chance at some point you may hit an issue caused by overclocking, even if it was rock solid in the past.

As for the 2nd part I highlighted, I'm not at all, and it isn't at all what I said and just goes to show you are not actually reading and understanding what is being said. You are using the "in my case everything works" scenario and aren't going beyond what docs told you on how to test for a stable overclock. If you don't understand after this, I can't explain it much better. I would suggest reading on how this all works together.

 

[ImpulsE69]

By all means, explain to me what happens when software "has an issue with overclock." I'll be getting my popcorn.

See above. If you are so certain, please explain to me how software is not impacted by overclocks, and how exactly you determine what a stable overclock is if you aren't using software?

For all the naysaying - not seeing anyone explain anything contrary, just a bunch of OMG NO MY OC IS STABLE nonsense. Also - I'm not overclocking, I actually underclocked from factory overclock which was pointed out as a solution to the problem on this game on cards that are factory overclocked (again in some cases). Try to keep up if you are going to chime in and try to contribute something.

 

[ImpulsE69]

Anyone with an AMD GPU that's Radeon 7XXX series or newer might wanna get this performance DLL file http://www.nexusmods.com/witcher3/mods/104

I tried it earlier, and I used the 1040 pack. At first I didn't think it was getting me any gains, but the increase was quite noticeable in some spots where I saw as much as 6-8fps higher. Such as spots that were 52fps without the DLL were pushed to 60. With all the settings maxed out, including hairworks with x8 AA, I can get 45-50fps running around Novigrad. That's with a 1130/1700 OC on my R9 390. This DLL is supposed to help in CPU intensive spots. So depending on the CPU you got, you might not see any gains, but an I5 2500K @ 4.2ghz bottlenecks occasionally on Witcher 3.

Thanks will check it out.

 

[cmdrdredd]

See above. If you are so certain, please explain to me how software is not impacted by overclocks, and how exactly you determine what a stable overclock is if you aren't using software?

For all the naysaying - not seeing anyone explain anything contrary, just a bunch of OMG NO MY OC IS STABLE nonsense. Also - I'm not overclocking, I actually underclocked from factory overclock which was pointed out as a solution to the problem on this game on cards that are factory overclocked (again in some cases). Try to keep up if you are going to chime in and try to contribute something.

All I see from you is "overclocking will eventually have a point of failure". That only happens if you go beyond what your hardware will allow. Every CPU, memory stick, GPU is different. You need to take your time and find what your hardware can and cannot handle. Also no, my system isn't going to reach a point of failure with some piece of software due to the clock speed or voltage. If it was going to, I'd have found out by now because I don't just click the Overclock button and walk away. I tune the voltages, clock speeds, calculate for vdroop etc all manually so that I reach the point I know anything I throw at my system will work correctly without issue. We're not talking about setting world records by going to the edge just to get a screenshot of superpi or 3Dmark. We're talking about overclocking a part to reach a good level of performance while keeping 100% stability. There is a big difference as I'm sure you are aware. There's reasons why running Prime95 for 24 hours or IBT (maybe both), memtest loops for 24 hours etc are done. They stress the components much further than any game or program you can run normally would. You can be sure that if your CPU runs 24hours or longer on Prime 95 and your memory checks out after a 24hour memtest run with no errors or crashes that a problem in a game isn't the result of an unstable overclock on your CPU and memory components. The GPU doesn't really have as many tools to loop for this testing but you can be pretty confident in your overclock if you can loop Heaven Benchmark, run Firestrike extreme a few times through and play your games for many hours with no artifacting or crashing etc.

So yeah I totally disagree 100% and I'm sure those over at extremesystems.com and overclock.net would to, that an overclock will always have a problem at some point. Only those who don't stress test have this issue. Again it is totally possible to have 100% system stability in an overclocked state.

 

[ImpulsE69]

All I see from you is "overclocking will eventually have a point of failure". That only happens if you go beyond what your hardware will allow. Every CPU, memory stick, GPU is different. You need to take your time and find what your hardware can and cannot handle. Also no, my system isn't going to reach a point of failure with some piece of software due to the clock speed or voltage. If it was going to, I'd have found out by now because I don't just click the Overclock button and walk away. I tune the voltages, clock speeds, calculate for vdroop etc all manually so that I reach the point I know anything I throw at my system will work correctly without issue. We're not talking about setting world records by going to the edge just to get a screenshot of superpi or 3Dmark. We're talking about overclocking a part to reach a good level of performance while keeping 100% stability. There is a big difference as I'm sure you are aware. There's reasons why running Prime95 for 24 hours or IBT (maybe both), memtest loops for 24 hours etc are done. They stress the components much further than any game or program you can run normally would. You can be sure that if your CPU runs 24hours or longer on Prime 95 and your memory checks out after a 24hour memtest run with no errors or crashes that a problem in a game isn't the result of an unstable overclock on your CPU and memory components. The GPU doesn't really have as many tools to loop for this testing but you can be pretty confident in your overclock if you can loop Heaven Benchmark, run Firestrike extreme a few times through and play your games for many hours with no artifacting or crashing etc.

So yeah I totally disagree 100% and I'm sure those over at extremesystems.com and overclock.net would to, that an overclock will always have a problem at some point. Only those who don't stress test have this issue.

If that is what you got out of what I said, you clearly aren't paying attention (or I'm horrible at explaining - which I won't rule out). We can agree to disagree, I'll take my decades of experience and work in IT and trust I know what I am talking about as it's done well by me. There are variables to everything. No amount of stress testing can account for all of it, and since that seems to be all you can tout, I am pretty sure your experience in the subject is fairly limited. I will not deny it is a small portion of issues, but then I never said it was some huge issue that you are trying to make it out to sound like.

 

[cmdrdredd]

If that is what you got out of what I said, you clearly aren't paying attention (or I'm horrible at explaining - which I won't rule out). We can agree to disagree, I'll take my decades of experience and work in IT and trust I know what I am talking about as it's done well by me. There are variables to everything. No amount of stress testing can account for all of it, and since that seems to be all you can tout, I am pretty sure your experience in the subject is fairly limited. I will not deny it is a small portion of issues, but then I never said it was some huge issue that you are trying to make it out to sound like.

Whatever makes you feel better.

 

[ImpulsE69]

Whatever makes you feel better.

I would feel much better if you'd quit talking like you thought I didn't know all the stuff you are already repeating over and over or that it pertained to me specficially (which it doesn't), and completely missing the points of my posts. Then I'd feel like we were having an actual discussion.

 

[Merad]

See above. If you are so certain, please explain to me how software is not impacted by overclocks, and how exactly you determine what a stable overclock is if you aren't using software?

So basically, you have no idea, hence the attempt to push the onus back on me to prove a negative.

A properly functioning processor is a deterministic machine. For any given set of inputs, it will always produce the same outputs. A malfunctioning processor (due to faulty hardware or overclocking) becomes non-deterministic. You expect a + b to equal c, but instead the processor hands you back z as a result. Pretty much by definition a processor with a "stable overclock" functions identically to a non-overclocked processor - i.e., correctly. The software doesn't change at all when the processor is overclocked.

If there's a statistically significant difference in the number of "bugs" you see when running a program on a OC'd CPU, then you can thank your unstable overclock.

 

[ImpulsE69]

So basically, you have no idea, hence the attempt to push the onus back on me to prove a negative.

A properly functioning processor is a deterministic machine. For any given set of inputs, it will always produce the same outputs. A malfunctioning processor (due to faulty hardware or overclocking) becomes non-deterministic. You expect a + b to equal c, but instead the processor hands you back z as a result. Pretty much by definition a processor with a "stable overclock" functions identically to a non-overclocked processor - i.e., correctly. The software doesn't change at all when the processor is overclocked.

If there's a statistically significant difference in the number of "bugs" you see when running a program on a OC'd CPU, then you can thank your unstable overclock.

Putting words in my mouth and assuming I don't know. You are the one who stated it was ludicrous. It is on you to explain your reasoning. I had already explained mine

I never said software changes because of overclocks in any way just like I never said an unstable overclock guarantees an issue. Again, people w/o reading comprehension. What does happen is software has expectations, and those expectations are met or/not met. I asked you what you use to determine a stable overclock for a reason. Software right? Are you going to determine it any other way? As long as the software gets what it determines is correct in terms of process/interactions it will function (as does the CPU and everything else). Stress tests serve as a good method for weeding out instability, however they are not absolute. What may work for one application may not work for another. There are 100's of documented cases of this happening since overclocking became a 'thing'. I still say that most of this would be covered under the case of "bugs", but again if out of 100 apps only 1 has an issue and removing/lowering the oc fixes it, the overclock is unstable. Just because you only have used 99 of those apps and haven't seen an issue, doesn't mean the possibility does not exist.

There is no way to 100% guarantee an overclock, period. Real world application is the true test of an overclock, not stress tests. As stated earlier they will provide you very good feedback and stability and in no way am I belittling the need for them, but in no way is it 100% all the time.

I do agree with your last statement "IF" removing that overclock solves your bug issue. Once again, your system could have run everything you ever threw at it flawlessly for years, but your system is only as stable as the last time it failed. I've never said an OC is a guarantee of failure, simply that the possibility will always exist, regardless of how stable you think your system is. Some applications are just touchy (or insert whatever you want here since we want to over analyze it).

All in all this discussion went off the rails because people around here have poor reading comprehension. As of yet no one has said why they think OC's are stable 100% of the time. Hardware alone and software alone do not make for stability. Out of spec is out of spec. That's why it's called out of spec. It may run awesome. Or it may not.

I do agree with cmdr that taking this to overclock.net could be interesting. I have no problems having an intelligent discussion about this. Unfortunately, that isn't what happens and you get responses like what we've seen here today. Also, it's hard to have these type of discussions w/o getting really wordy.

 

[escrow4]

You run all the software available that has built in checks. If it passes everything then it's stable. It's not that difficult and I don't know why you seem hell bent on proving that anyone who overclocks has an unstable machine.

The problem is that some programs are unstable to begin with which is why there are established and accepted standards for considering a system 24/7 stable. These accepted standards go way beyond any stress a game or working in photoshop or transcoding videos can do to a system.

ALL the software? You run a half dozen programs that stress the CPU/mobo and deem it stable? It ain't. An OC is never stable. Some software somewhere will dump it. Like the Witcher III. You may think its stable but it really isn't. You can never be certain a program you use won't knock it down.

 

[ImpulsE69]

What I'm noticing:

1. They are taking "oc is not stable" to mean - omg you're going to crash, quit overclocking

2. They think that stress testing solves all overclocking issues and means it is crash proof once they've ran it for however many hours they deem 'enough'

3. They think that software bugs can't be induced by overclocking (well this one is iffy. Merad states as much, he's saying "well your oc was never stable", but that's exactly what I'm saying too. It doesn't mean though that every program is going to crash. You may never ever (ever ever) see a crash. I think where we differ is the thought that there is or isn't an instance of an overclock that is 100% crash proof. Maybe not the best analogy, but just because you drive a car and never had an accident doesn't mean you can't no matter how many driver safety courses you take. The risk is always there.

 

[ImpulsE69]

Anyway - back to the actual topic discussion (that isn't really related to the above disagreement). I tried those dll's and still had the same crash to desktop issue but I agree, it did seem to give a bit of a boost in fps.

Just swapped back my old card and going to do some testing. It's not really a fair comparison and won't prove what the issue is (could simply be bad drivers). I pretty much already eliminated the clocking as my issue a day ago or so by underclocking the card.

It sucks because I took a break from this game and really am wanting to get back into it

 

[futurefields]

ALL the software? You run a half dozen programs that stress the CPU/mobo and deem it stable? It ain't. An OC is never stable. Some software somewhere will dump it. Like the Witcher III. You may think its stable but it really isn't. You can never be certain a program you use won't knock it down.

You could make this same statement about a stock clocked cpu/gpu then.

Some piece of software somewhere will dump it!

 

[Carfax83]

ImpulsE69, Merad and cmdrdredd all have valid points..

Overclocking is at once both very simple, and very complex mostly because there are so many permutations of hardware and software out there, plus not to mention ambient or environmental conditions such as cooling and airflow and differing chip limitations..

I'm not sure if there is ever anything such as a 100% stable overclock.. Let me give you an example..

Say I overclock my CPU to 4.6ghz, boot into Windows and then continue to run my repertoire of stress tests. Stress testing passes, I go play my games and I enjoy the extra performance the overclocking brings me without incident..

Then later on whilst browsing the internet, my computer freezes and I get a BSOD.

This incident can easily occur, and stress testing will not pick it up because stress tests only work for load states, and not idle..

In a situation like the above, what happens is that if you have the advanced C states enabled in the BIOS, the CPU will enter a very low voltage state during idle periods. But if the CPU is overclocked, the low voltage won't be enough to sustain operations and it will BSOD during innocuous desktop usage.

The only way to deal with that from my experience, is to either disable the C states entirely (and deal with much higher voltages during idle usage), or enable a defined level of load line calibration to prevent the voltage from dropping too low.

I usually do the latter, as I want at least C1 enabled because it's a good feature. But my point is that overclocking is very rarely ever cut and dry, and there's tremendous potential for things to go awry. This escalates dramatically the further out of spec you push the hardware..

 

[cmdrdredd]

ALL the software? You run a half dozen programs that stress the CPU/mobo and deem it stable? It ain't. An OC is never stable. Some software somewhere will dump it. Like the Witcher III. You may think its stable but it really isn't. You can never be certain a program you use won't knock it down.

You could name a hundred programs that crash for people and there would be innumerable reasons for it to possibly happen.

 

[cmdrdredd]

ImpulsE69, Merad and cmdrdredd all have valid points..

Overclocking is at once both very simple, and very complex mostly because there are so many permutations of hardware and software out there, plus not to mention ambient or environmental conditions such as cooling and airflow and differing chip limitations..

I'm not sure if there is ever anything such as a 100% stable overclock.. Let me give you an example..

Say I overclock my CPU to 4.6ghz, boot into Windows and then continue to run my repertoire of stress tests. Stress testing passes, I go play my games and I enjoy the extra performance the overclocking brings me without incident..

Then later on whilst browsing the internet, my computer freezes and I get a BSOD.

This incident can easily occur, and stress testing will not pick it up because stress tests only work for load states, and not idle..

In a situation like the above, what happens is that if you have the advanced C states enabled in the BIOS, the CPU will enter a very low voltage state during idle periods. But if the CPU is overclocked, the low voltage won't be enough to sustain operations and it will BSOD during innocuous desktop usage.

The only way to deal with that from my experience, is to either disable the C states entirely (and deal with much higher voltages during idle usage), or enable a defined level of load line calibration to prevent the voltage from dropping too low.

I usually do the latter, as I want at least C1 enabled because it's a good feature. But my point is that overclocking is very rarely ever cut and dry, and there's tremendous potential for things to go awry. This escalates dramatically the further out of spec you push the hardware..

If you want to use low voltage idle states you test for that as well and make sure your voltage doesn't drop below a level that keeps stability. It goes hand in hand with full load testing. That's how I handle it anyway because I too use the lower voltage c states.

 

[escrow4]

You could make this same statement about a stock clocked cpu/gpu then.

Some piece of software somewhere will dump it!

In that case its not the overclock it could be hardware or something else. Doesn't change the fact that you never know if an OC is stable.

 

[Artorias]

For what its worth I've been using a factory OC R9280x and have had very few crashes, in fact I think I've only crashed during cutscene transitions which was a bug that was later fixed.

I'm not that informed on overclocking, but how would a program differentiate if your card is "overclocked" and your not just using another GPU with a faster standard clock rating. Just speculation here, but wouldn't crashes more so have to do with the GPU mechanisms not being in sync(not rated to go faster) because of an overclock, like several gears all of a sudden cant keep up causing GPU hang.

 

[Chaotic0ne]

I've pretty much accepted that there is going to be the occasional artifact in Witcher 3, but I got it down good enough where I can go 30 minutes or an hour without seeing one, and when I do see one, I don't see another one for awhile. It could very well be a regular graphics glitch, because some of them I'm seeing at stock clocks as well. Every now and then then there is a large black rectangle that is about the size of Geralt that will appear to his left or right side. I'm convinced 100% its the game, or the drivers causing this. FYI Witcher 3 has never crashed on me once. Been playing the game from 1.01 up to 1.08. And I've played the game in 2 hour sittings many times. If better FPS means seeing an artifact once every 30 minutes for .1 secs, and the game doesn't crash on me then that's a good trade IMO. That's my definition of stable enough.