|
On October 02 2013 06:34 Torte de Lini wrote:
What's the best way to have a vertical and horizontal monitor setup? I know its preference in the end; but I'm trying to figure out how to enjoy games without going off the screen when edge-panning (like in Dota 2), but also maintain a comfortable flow of dual-monitors (one that is vertical and the other is horizontal).
Not sure if I'm being clear here :x
Most modern games have the confine mouse cursor option so you won't be going off the game screen unless you have some retarded program installed that has popups that alt tabs you.
Not sure what you're asking for. You either use the adjustable stand or a VESA mount.
|
Strange, dota 2 doesnt under any circumstances (fullscreen/borderless window, etc.)
|
If you play in borderless window, the mouse will stay in the game.
|
fixed but borderless doesnt create exclusive windows/mouse stays in game, this was obs mode though
|
Idk than, I know borderless window worked back when I was playing.
|
PrimaryLock is a hackish fix for that.
|
Is it normal for CPU overclock to hit asymptotic limits (I'm a total script kiddie noob at this). Re my setup here + Show Spoiler +ASrock Z77e-ITX
i5 3570k
MSI 7950 (the 7970 PCB batch)
ADATA 2x 4GB DDR3-1600 CL 9
I figured out how to get everything to fit (I used zip ties to mount the GPU radiator in between my hard drive cage and the top radiator to the front of my case since the default screw mounts were in undesirable locations) and so I wanted to push my limits. The problem is that my outputs are completely nonlinear given my desired clock.
I.e. @ 43x, base level vcore offset, base level turbo I'm at 67C after an hour of OCCT:linpack
@44x, still stock voltages, 71C
Attempting 45x it seems I just hit a hard ceiling. turbo vcore +.008 through +.016 have all blue screened, and at +.020 I'm hitting 90C and getting rounding errors fatal error in prime95 and object sharing errors in linpack.
|
Not all CPUs are equal so yes. 70C is normaal but turbo vcore doesn't really tell us anything. It's important that you are monitoring your vcore when you are stress testing your overclock.
|
@34 Vcore converges 1.216
@44 Vcore converges 1.312
@45 Vcore converges 1.36
I'm mostly asking if this kind of behavior is typical. From 34 up to 44 each increase in multiplier saw a proportionate increase in power draw and peak temps. @45 everything just explodes.
|
Yeah, pretty much.
Probably somehow involved is the way in which heat and electricity loss is exponentially related.
|
Hm? Heat and power consumption (electricity loss) should be pretty much almost the same, so linearly related.
Given that you're fully loading the thing, power consumption would be expected to be proportional to the square of the voltage if it were a linear system, but it's not one in reality. Modeling everything going on inside would be nuts. But anyway, increased voltage also means higher temperatures, which also increases power consumption in of itself. You can expect a lot of the relationships to not hold at certain points along the curve though, because of the nonlinearity and weird factors involved.
Anyway, the difference in voltage (1.36V vs. 1.312V) between the settings for 4.5 GHz and 4.4 GHz would have a lot to do with it. I'm guessing you didn't use a big a voltage difference between 4.4 GHz and 4.3 GHz? That's why things are running away from you.
According to what is experienced by most overclockers, yes, it seems like empirically people run into a cliff like that on most samples and systems.
|
United Kingdom20344 Posts
0.048v for 100mhz is not too bad, it's just not a great chip if you're needing that much vcore and difference by 45x
|
I got it stable through hour of linpack and prime. 4.5 GHz @ 1.392 V, temps spike up to like 88 for first 15 mins then drop to 70s (might be autocontrol on my fans or something). Probably not gonna bother trying 46x since it seems like I'm pretty much at the asymptotic cliff. Thanks for the info guys.
|
On October 02 2013 09:50 xes wrote:
@34 Vcore converges 1.216
@44 Vcore converges 1.312
@45 Vcore converges 1.36
I'm mostly asking if this kind of behavior is typical. From 34 up to 44 each increase in multiplier saw a proportionate increase in power draw and peak temps. @45 everything just explodes.
A setting you may not have played around with is the LLC setting (Load-Line Calibration). You can find it if you scroll down to the bottom on exactly that BIOS screen you are looking at normally when overclocking. Use a setting that's somewhere in the middle, not the lowest and not the highest. That's what is usually best for offset voltage overclocking.
It tweaks how high the voltage goes when the CPU uses a lot of power. On ASRock boards, 0% or level 0 will mean the highest voltage while running prime95. You will likely have to tweak the turbo offset voltage after you change LLC.
+ Show Spoiler +
There's "fixed voltage" on other boards, but ASRock does not have that. It always runs with dynamically changing voltages like defined by Intel to save power and keep heat in check.
The CPU demands a certain voltage through the "VID" signal. The VID signal is not fixed, it changes depending on the instructions being processed. There's not really a single "stock voltage". Every CPU is unique.
To enable overclocking, the offsets you've set in the BIOS get added and the result is the vcore.
Another rule is called "vdroop". The more power the CPU uses, the larger vdroop gets. Vdroop gets subtracted from vcore. LLC is a factor for vdroop.
LLC will influence how high the voltage goes when the CPU uses a lot of power. You can try to curb the voltage you'll see for stress tests specifically. If you pull down the voltage with LLC, you will likely have to increase your turbo offset to get back to stability.
On ASRock, the LLC setting works in reverse compared to ASUS etc. If you choose the lowest setting (level 0 or 0%), you get max LLC, vdroop will be eliminated (or even reversed). If you choose the highest settings (might be something like level 3 or level 5 or 100%), LLC will be completely off, vdroop will not be touched.
When prime95 runs, the CPU will set a very high VID value. It expects vdroop. If you eliminate vdroop completely, you will get very high voltage and make testing stability with prime95 a bit pointless. That's why it's best to use some LLC setting in the middle with offset overclocking. Fixed voltage, you'd want a flat voltage at all times and would set LLC to have zero vdroop.
|
I have a TL mousepad and wondering if anyone knows how to clean it?
|
If it's a bit like the cloth pads I've used (Allsop Raindrop and Steelseries QCK+), you can soak it in the sink in warm water and go at it with shampoo and nothing bad will happen to it. Drying takes a while.
|
United Kingdom20344 Posts
^I also just wet a cloth/towel thing with warm water and squeeze it out so it's lightly damp a couple times a day for cleaning mouse/kb/mousepad/hands etc. Maybe a bit high maintenance for some people but it seems to work. Never washed my qck+ with shampoo or washing machine, maybe i should try it 
|
United Kingdom20344 Posts
Hey, is there any free or easily available dx11 game that is somewhat demanding (doesn't have to be crazy, just an FPS cap of 200 and always there without drops on max settings is quite useless) and has any tightly reproducable benchmark? I might borrow crysis 3
|
On October 03 2013 11:19 Cyro wrote:
Hey, is there any free or easily available dx11 game that is somewhat demanding (doesn't have to be crazy, just an FPS cap of 200 and always there without drops on max settings is quite useless) and has any tightly reproducable benchmark? I might borrow crysis 3
What are you plotting now cyro? 
|
United Kingdom20344 Posts
Opening people's eyes on streaming performance hits! 
Within 5 months i should have a nice monitor and ~30/6 internet, hopefully! It will be more relevant for me then, but AFAIK dx9 is much more painful than dx11 in terms of losses before taking encoding CPU load into account. Not seen enough numbers, so i'd like some myself
Why the face? :0
|
|
|
|
|
|
EPT
