Witcher 2 putting my CF 7970 onto its knees? - Page 11

Did you check your CPU usage for any of those games? Find out how many cores each of those games can use then check your core usages. If the game uses only 2 cores and the two cores the game is using are not maxxed then the CPU is not a bottleneck. It's really not that hard.

Dude, you're the one that needs to take the breath. My post is perfectly calm.

And 'slander'? Oh, please

You are the one who was incorrectly expressing the idea that <100% CPU usage means no bottleneck ... don't blame me for clarifying your inaccurate information, which leads to misunderstandings among people coming here to learn. You think everyone who comes on here knows how to interpret your statement properly? I know for a fact ... they don't. It's kinda my job to clarifying when people state things in an inaccurate way.

Also, I did my own Witcher 2 tests nearly a year ago, trying to reckon why it was that I frequently saw low fps and low gpu usage in certain areas. Ergo, I know for a fact that this exact condition at least used to occur with other than AMD drivers, because I was running my sig rig (SLI 470's, i7 930 at 4.0GHz) at the time.

One thing that I discovered in my testing is aptly illustrated by the following screenie, which like I say I did nearly a year ago.

Now, when I ran this test, it was with everything in the game at ALL LOW, no AA, at 1024x768 (or maybe it was 800x600 ... whatever is lowest in the game). FPS was plenty high ... in the 80-120 range as you can see. Observe where the CPU usage lies. And the GPU usage, for that matter. Does it not appear as though I've created a CPU bottleneck in this scenario, at a mere 25% of CPU usage?



Now, I really don't care what people say on the internet in some review when my direct observations fly directly in the face of what they're telling me. It's very clear from my cpu usage that this game is basically a two core game, which can maybe use some resources of a 3rd core. That's why usage caps out around 25-30% overall even when I've done everything possible to max out CPU usage.

Hence, I have pretty compelling evidence to back up exactly what I said. I don't NEED 'google'

Now keep in mind I'm by no means saying that OP's problem MUST be a CPU BN. I am, however, clarifying that not only AMD has had this issue in this game, and that, generally-speaking, it most certainly IS possible to be CPU BN'd at <100% cpu usage on a multi-core CPU.

To me, the OP's description of the problem (as well as my own observations from last year) reminds me quite a bit of the problems that Skyrim had for it's first 4 months in certain areas of the game (i.e. the steps at Dragonrun), before Bethesda patched it with proper CPU optimizations.

Dude, for the Witcher 2 on a 2600k anything less than 100% CPU usage means no CPU bottleneck. There's nothing to interpret. It uses all the cores including hyperthreading. Look it up... I guess the developers are lying too.

You have posted nothing to indicate a CPU bottleneck. In fact, the highest point on your usage graph is exactly that, a point. Do you even know what a CPU bottleneck is or what it looks like on a graph? You also do not show the load per core either.... For all I know from what you posted, that's a single core processor... I'd hardly call that proof of anything.

Do I even know what a CPU BN is or what it looks like on a graph ... stop ... you're killing me

Frankly, if you don't have enough basic understanding to know that this particular CPU and GPU usage graph, under the conditions I described (all low, no AA, 800x600) clearly illustrates a CPU-bottlenecked condition caused by a game only using 2 threads (out of the 8 available ... what you think I'm LYING about my proc? Oh, please) ... then I'm not going to be the one to take the time to explain it to you. I've explained this phenomenon so many times on this board that I can't be arsed to do so again.

Also ... individual core usage in Task Manager doesn't tell you anything useful except what Windows would LIKE to do with the load. It's not a bare-metal measurement of actual core activity, and as such it doesn't necessarily accurately tell you how the load is being handled by the processor.

http://technet.microsoft.com/en-us/library/cc750967.aspx

I'm done.

Dave, I understand your confusion ... you've 'been told' that the game can 'use' all cores, all threads, HT, and all that. So you're taking what you've been told at face value, and then extrapolating that to mean '<100% usage means no CPU BN'. That's understandable, I suppose.

However, I'm telling you, as someone with a pretty high level of knowledge of how to go about such things ... I've tested this game ... and it is fundamentally a dual core game that can occasionally use some of a third core.

Furthermore, the following review at least somewhat backs up what I'm telling you:



http://benchmark3d.com/the-witcher-2-benchmark/2

And again, I'm not saying the problem IS the CPU here, just to be clear. I'm merely stating that one cannot rule out a CPU BN simply because usage is <100% in this, or frankly in any other game. And I'm also saying that I most definitely had the same problems with my nV cards as what the OP describes, so there is at least some doubt as to whether this problem can definitively be pegged on AMD's drivers. It may BE AMD's fault, but I'd have to see more evidence to be convinced. I'll be doing some testing of my own once the 10GB d/l is finished in order to see how the game (and nV drivers) have progressed since June 2011, when I finished playing it.

I will also point out that my testing was done specifically in the 'troublesome' parts of the game (i.e. inexplicably low GPU usage, when the FPS is also low), and as such it may well be that the 'troubles' are actually being caused by game only being able to use two cores in these specific areas. It's not uncommon for games to only exhibit CPU BN at certain spots ... see Skyrim in it's first 4+ months of existence for evidence thereof

What have you done exactly?

Showed them how to properly identify a CPU bottleneck. Did you not read any of it?

What is the thread queue like while running the game? Unless the thread queue is constantly greater than 0 then there is no CPU bottleneck. A bottleneck occurs when the CPU cannot process the threads as fast as they are being queued. Starting more processes will only make the queue even greater at that point. Under most normal circumstances the Processor Time will be 100% (90%+) before there is a thread queue greater than 0 unless multiple threads bombard the processor at once.

Requests made from the GPU are threads. A CPU bottleneck in a game occurs when the GPU has to wait for the CPU to process the threads it requires. If the thread queue is not greater than 0 then the GPU is not waiting on the CPU.

This is one example of what a CPU bottleneck looks like:


The red line is the thread queue. The green line is the processor time.

The red line never drops to 0 and the processor time is maxxed. This is exactly why the "per core" load is very important. But the thread queue is more important. Showing the total processor time of a multi-core processor means nothing unless the process is utilizing all cores (which the mod says it doesn't). It also means nothing without showing the thread queue. No waiting line in queue = no bottleneck. No matter what the processor time looks like.


I was trying to avoid a long a detailed argument about this by providing the info for everyone to read themselves. If you want to prove a CPU bottleneck, post a thread queue graph too. If the bottleneck is only apparent in specific areas, then the thread queue graph will also reflect that.

Good information, thanks Dave. Unfortunately I can't go back a year in time to check the processor queue, so we'll just have to see once my d/l is done whether the previously existing issue has been corrected.

However, to address this point:
The way that one MAKES the total processor time 'mean something' in a gaming scenario is by lowering all settings to their lowest, including resolution. By doing this, the GPU load is minimized, FPS will skyrocket, which is turn maximizes demand on the processor. This creates a very high probability that the CPU will become the limiting factor to performance.

If, at this point, one observes total CPU usage, and see's it basically 'pegged' at 25% (i.e. a nice, even multiple of 12.5% on a 8 threaded system), ALONG WITH the GPU usage NOT being 99% or thereabouts, it provides an extremely compelling case that the game is only capable of using 2 out of the 8 available threads (i.e. 25%), at least in the portion of the game where the testing is being done. Granted, it's not 100% certain, but this is by far the most likely explanation.

I'll leave it to you Dave to provide a compelling counter-explanation for why, in the scenario I described, my GPU usage is nowhere near 99% ... if mine is not the correct interpretation of what's going on there

Lastly, Dave, I think at least some of the generalizations you're making may be dependent upon one of these two being true:
1) you have a single core processor, or
2) The application is coded to be able to use all available threads at 100% capacity.

Things become murkier in the world of gaming where games are coded for specific numbers of threads.