EPTThe Ultimate Mouse Thread - Page 23
| Forum Index > Tech Support |
|
Medrea
10003 Posts
| ||
|
IPS.Blue
Germany309 Posts
On January 17 2012 21:24 Medrea wrote: If 1k polling is dropping FPS you have bigger problems. Don't underestimate that. I have a fast dual-core and I can watch the CPU usage spike when moving a 1000 hz mouse with nothing running but a browser. Now imagine running SC 2 on a laptop at a LAN ... And even if you have the "power", many prefer 500 over 1000 hz because of more stability. If your hz change something like this: ![[image loading]](http://i44.tinypic.com/ak7m3c.jpg) you maybe sometimes have a faster response time or less prediction, but on the other hand, you have less consistency. | ||
|
Medrea
10003 Posts
You are probably just mousing over elements on the screen which causes the MOUSE_OVER response to happen rapidly... You can reproduce with desktop icons. Mouse over empty space and nothing major will happen. Mouse over icon clutter and something will happen. I cant even get my CPU to stop downclocking itself with just the mouse and Recorder. I mean really, we have been mousing around since 33mhz processors were the shit. And that was from the PS/2 which is a direct hardware interrupt. Your Movement recorder is normal, I dont think you understand how polling works. There is no instability. Recorder will not report a null value,and doesnt report the poll taking place until the next successful poll. Polling rate cannot "change" or anything like that. Its constant. If it takes 2 polls to get a value, 1000/2=500. If you keep your mouse still Recorder stops because it will not display a value of 0 x 0. Same thing happens when the mouse doesnt update fast enough. 500hz is always less "stable" than 1khz. But thats a really silly way of looking at it. Players with weak PC's choose 500hz because they cant do 1k or the motherboard will freeze. "Asking" for a value from the mouse when it isnt ready doesnt fuck anything up. It is just a poll. Its not like your motherboard is trying to rip the bits from the mouse. | ||
|
IPS.Blue
Germany309 Posts
On January 17 2012 21:57 Medrea wrote: Moving the mouse does not incur hardly any processing power. Its running on a different hardware layer than everything else -.- You are probably just mousing over elements on the screen which causes the MOUSE_OVER response to happen rapidly... You can reproduce with desktop icons. Mouse over empty space and nothing major will happen. Mouse over icon clutter and something will happen. I cant even get my CPU to stop downclocking itself with just the mouse and Recorder. I mean really, we have been mousing around since 33mhz processors were the shit. And that was from the PS/2 which is a direct hardware interrupt. I have no items on my desktop. My CPU usage normally is at 1-2 % (sometimes spikes up to 20; 2x 3,33 GHz). When I start moving my mouse it spikes up to 50-60 shortly after (although it drops relatively fast). Anyway, there is not much use in debating whether 1000 hz drains more CPU power or not - more polling is simply more work for your CPU. The question is: how much? How would one be able to measure something running on "a different hardware layer"? On January 17 2012 21:57 Medrea wrote: ... "Asking" for a value from the mouse when it isnt ready doesnt fuck anything up. It is just a poll. Its not like your motherboard is trying to rip the bits from the mouse. I didn't say the mouse would die. If you poll your mouse at a high rate, but your mouse can't always deliver, you end up with inconsistency. Experienced players are able to feel the difference between 125, 500 or even 1000 hz. Ergo will an experienced player note the difference between 1000 hz all the way and 333-1000 hz. | ||
|
Neurosis
United States893 Posts
| ||
|
Deleted User 135096
3624 Posts
| ||
|
Wintex
Norway16835 Posts
 Win~ | ||
|
TT1
Canada9987 Posts
On January 18 2012 06:38 Wintex wrote: I am keen on buying the G9x. How is it to switch to it from the MX518, Deathadder and Diamondback? Win~ horrible? completly different grip, ud most likely hate it if u can get used to tho it its one of the best mice on the market atm | ||
|
Wintex
Norway16835 Posts
On January 18 2012 06:45 mTwTT1 wrote: horrible? completly different grip, ud most likely hate it if u can get used to tho it its one of the best mice on the market atm <3 TT1! Ah well, let's put the nerdchills aside, thanks! I really "attune" to mouses easily (except the Razer Naga for some reason). If it is that good I'll go for it! Edit: T.T grammar sux Win~ | ||
|
Medrea
10003 Posts
On January 18 2012 02:57 Neurosis wrote: But does it actually cause inconsistency if it's bouncing between low and high hz. I haven't really felt any inconsistencies since patching my mx518 to 1k and it has definitely been bouncing around something ridiculous. OK I think I need to explain what polling rate is and how mice work because I am obviously not being clear somewhere. When a mouse detects motion it uses a high sensitivity sensor to detect how much it is moving. The sensor takes pictures many times a second. An MX518 takes about 700 or so pictures a second and todays mice I think a little over 1,000 pictures a second. The mouse compares pictures every single time and measures the difference between the two based on granular differences in the mousing surface. After doing some math the sensor coordinates the difference into a unit called the DPI. Or Dots Per Inch. If you have a 2 DPI mouse, and the mouse sees you moved it one inch to the right, it will post 2 x 0 into the mouses output register. If the motherboard happens to poll between now and the next time the mouse updates that register. The mouse will first answer the poll, and then clear the registry. If the motherboard does not poll the register in time, the mouse will keep the value in the register, and add the next one on top of it. So if the motherboard is late you will post 2 x 0 and another 2 x 0 if you keep moving the mouse, resulting in the second updated value being a 4 x 0. If the motherboard polls twice, the first poll takes the value from the mouse, and then the mouse clears the register. The second poll happens before the mouse has taken a picture. The returned value is NULL, and the poll is discounted. It happened, but the value is not acceptable. The poll still happens. The reason why you are seeing common values of 1000, 500, 333, 250 and so forth (look at the movement checker above on this page) is because the motherboard polled 1 2 3 and 4 times before a value was found. As: 1000/1 = 1000 1000/2 = 500 1000/3 = 333 And so forth. Only valid data returning presents a valid poll. There is no "bouncing." That is just the user misinterpreting the tool. And mice have been in use forever and has never in all PC time been a significant CPU power draw. 20+ years later eight times zero is still zero. If FPS.Blue is experiencing problems, thats his PC. In reality he is probably mousing over a tooltipped element on screen without realizing it. No, we dont need powerful dual cores to render a mouse -.- a first gen Pentium I does a fine job. If you have a problem with this, you should be plugging the mouse in the PS/2 port then. Even USB mice operate on a different hardware layer, thats why your mouse works even though the rest of your system might be dead. PS/2 is actually literally a different hardware layer since it is a direct hard interrupt on the system. Polling a mouse 500hz introduces latency of up to 2ms. I say "up to" because you really have no idea when the mouse is going to get polled between 0ms and 2ms of the register being updated. It could be 0.1 ms, it could be 1.1 ms. Polling at 1k introduces up to 1ms. Thus, if you can poll at 1k, poll at 1k. 1k or bust. In fact 500hz is less accurate because of the remainder effect I described above. Consistency is not a word here unless you are saying you can actually predict when the motherboard is about to take a poll, in which case I would call extreme bullshit. There is only less, and more accurate. A Mouse thats being read 1,000 times a second and sometimes 500 is always more accurate than a mouse thats just being read 500. And for keyboards, keyboards have a debounce time of around 5ms. Thus, polling a keyboard faster than 200hz is useless. And any keyboard maker (I am looking at you Thermaltake) that markets a faster polling keyboard is purely fluffing you. | ||
|
TT1
Canada9987 Posts
On January 18 2012 07:05 Wintex wrote: <3 TT1! Ah well, let's put the nerdchills aside, thanks! I really "attune" to mouses easily (except the Razer Naga for some reason). If it is that good I'll go for it! Edit: T.T grammar sux Win~ haha ur so cute<3, dunno man the decisions up to u.. the mouse itself is very good but im worried that u wont get used to the size of it : S, its very small | ||
|
snexwang
Australia224 Posts
Also, I'd recommend the CM Spawn over the G9X if you're using a black mousepad. Look that shit up. :> | ||
|
Zurachi
Canada289 Posts
| ||
|
Khenra
Netherlands885 Posts
| ||
|
Deleted User 135096
3624 Posts
On January 18 2012 19:34 Khenra wrote: I tried the G9x but it's really, really heavy. It's really small and really heavy, which makes it very awkward for me to use. And that is without any of the weights in it.. have you tried it without the grips? The base is only about ~90g without them, ofc the side buttons I've heard are harder to press without the grips. | ||
|
IPS.Blue
Germany309 Posts
On January 18 2012 07:14 Medrea wrote: OK I think I need to explain what polling rate is and how mice work because I am obviously not being clear somewhere. When a mouse detects motion it uses a high sensitivity sensor to detect how much it is moving. The sensor takes pictures many times a second. An MX518 takes about 700 or so pictures a second and todays mice I think a little over 1,000 pictures a second. Optical sensors reach about 6000 (the MX518 with up to 6469) and laser sensors scratch the 12000. Plus there are sensors who work completely different (Philips Twin Eye). On January 18 2012 07:14 Medrea wrote: The mouse compares pictures every single time and measures the difference between the two based on granular differences in the mousing surface. After doing some math the sensor coordinates the difference into a unit called the DPI. Or Dots Per Inch. It's CPI. DPI is just plain wrong. A printer has DPI. A mouse, as you said, counts, it doesn't paint dots. On January 18 2012 07:14 Medrea wrote: If you have a 2 DPI mouse, and the mouse sees you moved it one inch to the right, it will post 2 x 0 into the mouses output register. If the motherboard happens to poll between now and the next time the mouse updates that register. The mouse will first answer the poll, and then clear the registry. If the motherboard does not poll the register in time, the mouse will keep the value in the register, and add the next one on top of it. So if the motherboard is late you will post 2 x 0 and another 2 x 0 if you keep moving the mouse, resulting in the second updated value being a 4 x 0. If the motherboard polls twice, the first poll takes the value from the mouse, and then the mouse clears the register. The second poll happens before the mouse has taken a picture. The returned value is NULL, and the poll is discounted. It happened, but the value is not acceptable. The poll still happens. The reason why you are seeing common values of 1000, 500, 333, 250 and so forth (look at the movement checker above on this page) is because the motherboard polled 1 2 3 and 4 times before a value was found. As: 1000/1 = 1000 1000/2 = 500 1000/3 = 333 And so forth. Only valid data returning presents a valid poll. There is no "bouncing." That is just the user misinterpreting the tool. ... Polling a mouse 500hz introduces latency of up to 2ms. I say "up to" because you really have no idea when the mouse is going to get polled between 0ms and 2ms of the register being updated. It could be 0.1 ms, it could be 1.1 ms. Polling at 1k introduces up to 1ms. Thus, if you can poll at 1k, poll at 1k. 1k or bust. In fact 500hz is less accurate because of the remainder effect I described above. Consistency is not a word here unless you are saying you can actually predict when the motherboard is about to take a poll, in which case I would call extreme bullshit. There is only less, and more accurate. A Mouse thats being read 1,000 times a second and sometimes 500 is always more accurate than a mouse thats just being read 500. ... This exceeds my technical knowledge. Do you have sources or is that stuff just floating around your brain  ?Your example is flawed, though. If a mouse takes 12000 pictures a second and gets polled 1000 times a second - how can it be, that it can't deliver new data (when being moved rapidly in circles)? Also mind, that the serial port of the sensor often operates at 1 or 2 MHz ... | ||
|
DystopiaX
United States16236 Posts
On January 18 2012 19:34 Khenra wrote: I tried the G9x but it's really, really heavy. It's really small and really heavy, which makes it very awkward for me to use. And that is without any of the weights in it.. Yeah it's not for everyone, but personally I like a solid, really heavy mouse so I put the max weight in it. The R.A.T 7 and the new Thermaltake mouse introduced at CES are interesting in that they're hollowed out in the middle, and so they're really light unless you put the weights in. | ||
|
MrBazinga
32 Posts
On January 19 2012 06:57 IPS.Blue wrote: 1. Optical sensors reach about 6000 (the MX518 with up to 6469) and laser sensors scratch the 12000. Plus there are sensors who work completely different (Philips Twin Eye). 2. It's CPI. DPI is just plain wrong. A printer has DPI. A mouse, as you said, counts, it doesn't paint dots. 3. This exceeds my technical knowledge. Do you have sources or is that stuff just floating around your brain ?Your example is flawed, though. If a mouse takes 12000 pictures a second and gets polled 1000 times a second - how can it be, that it can't deliver new data (when being moved rapidly in circles)? Also mind, that the serial port of the sensor often operates at 1 or 2 MHz ... It's late and I'm tired, but I'm trying to get to the point: 1. He was talking about how many pictures per second are being taken, you are talking about CPI which is, so to speak, the resolution of the sensors. 2. We all know it means CPI. Yet DPI is an accepted term that even the manufacturers put on their retail boxes. It's like with cars and PS vs kW. 3. Well if the mouse has nothing to report, it has nothing to report, I'd say. This also exceeds what I can say I know for sure about that stuff, but it sounds very logical to me and actually like the only reasonable explanation I could think of. (and again, you are mixing up the numbers of "how often does the sensor scan" and "what is the accuracy (or resolution) of these scans".) | ||
|
Medrea
10003 Posts
On January 19 2012 06:57 IPS.Blue wrote: Optical sensors reach about 6000 (the MX518 with up to 6469) and laser sensors scratch the 12000. Plus there are sensors who work completely different (Philips Twin Eye). It's CPI. DPI is just plain wrong. A printer has DPI. A mouse, as you said, counts, it doesn't paint dots. This exceeds my technical knowledge. Do you have sources or is that stuff just floating around your brain ?Your example is flawed, though. If a mouse takes 12000 pictures a second and gets polled 1000 times a second - how can it be, that it can't deliver new data (when being moved rapidly in circles)? Also mind, that the serial port of the sensor often operates at 1 or 2 MHz ... That number is different from the actual rate at which the buffer gets updated. You are actually talking about sample rate of the sensor, I am talking about how often the value in the mouse gets updated after the data is processed through the mouse firmware and all that. Its good you mention that though. Sample rate of the sensor is huge when it comes to talking about the max speed of a mouse. And I agree, I was somewhat economical with that part of the explaination. I dropped a big knowledge bomb so admittedly tracking down the sources for absolutely all of that is tricky. Even those eHow websites really boil it down. However, I might add that actually all devices since USB was conceived work in approximately the same way. If you still think that 1k and sometimes 500 is unstable, consider a circle (circle movement). 500hz straight up is going to give you a certain amount of data points within that circle. If I just add more data points in between, I can only increase the resolution and accuracy of the circle I made with my hand. The points could be completely random, it doesnt matter, more points of data is always better. I could add just one, one single additional data point, and it would be better. Unless of course you are "banking" on the polling rate delay in which case you would be the motherfucking Rainman. I appreciate your sentiment about DPI. It is a bit of a misnomer. | ||
|
lisward
Singapore959 Posts
| ||
| ||
