Overclock.net banner
1 - 20 of 59 Posts

·
Mouse addict
Joined
·
795 Posts
Discussion Starter · #1 · (Edited)
Disclaimer: Although the review sample was provided by Endgame Gear, the review itself is 100% unbiased. All opinions are my own.

Features
The Endgame Gear XM1 is an ambidextrous gaming mouse with side buttons on the left side only (six buttons in total) that comes with Pixart’s latest top optical sensor, the PMW3389, which is capable of up to 16,000 CPI. Furthermore the XM1 features a lightweight construction of just 70g, 100% PTFE mouse feet, pre-selected Omron switches for the main buttons (rated for 50M clicks) and lastly a patented analogue switch contact technology, which promises a click delay of <1 ms without any unwanted double click actuation. A button for on-the-fly CPI and polling rate adjustment on the bottom along with software functionality are present too.




Packaging
The XM1 comes in a simple yet solid box, with the mouse sitting nicely cushioned in a foam package. The cable is somewhat tightly folded, so it may take a while until the kinks disappear. There are no replacement mouse feet in the box, but the provided ones should already last a while. Endgame Gear will also sell replacement feet.





Dimensions:
Base length: 11.3 cm
Length including overhang: 12.2 cm
Height (highest point): 3.6 cm
Front height (lowest point): 1.3 cm
Width (widest point): 6.6 cm
Grip width (narrowest point): 5.5 cm





Weight
My scale returns a weight of 69g, which is 1g less than the proclaimed 70g weight.



Weight balance is very good as well, which makes the XM1 feel even lighter than it already is.



Comparison: Steelseries Kana at the top, Nixeus Revel at the bottom

Shape
The shape of the XM1 has been co-developed by CS 1.3/1.5/1.6 AWP legend Johnny R, who also had a hand in designing the Steelseries Sensei, Kana and Kinzu shapes. Hence it comes as no surprise that the shape of the XM1 is rather close to that of those three mice. As you can tell from the comparison pictures, the XM1 is actually most similar to the Kana, with the greatest differences being a slightly more tapered butt, the hump being more towards the back and the XM1 being slightly shorter and wider. The uniquely curved sides along with the widened butt are strongly reminiscent of the Sensei and Kana. Since I’m a big fan of the Kana the XM1 is right up my alley shape wise. The feel in hand isn’t exactly the same as the Kana though, but quite close.
As for grip styles, I’d say the XM1 should be suitable for all three grip styles as well as most hand sizes (it’s a medium size mouse). However, I reckon that it’s most suitable for claw and fingertip grip and least suited for palm grip. Although there is some palm support due to the hump being positioned towards the back, it’s too flat to provide full support.




Comparison: Microsoft WMO at the top, Zowie FK2 in the middle, Dream Machines DM1 Pro S at the bottom

Mouse feet and cable
The feet are made of 100% PTFE, which is why they’re white instead of the usual dyed black. 100% PTFE feet should result in greater longevity compared to non-100% PTFE feet. Glide is excellent on the cloth pads I tested it on. There are two medium (at the back) and two small (at the front) feet, the latter of which could be difficult to replace with generic feet. Since Endgame Gear will provide fitting replacement feet this is a rather minor complaint, however.
The cable is a standard rubber cable, which is not as flexible as the Model O or DM1 FPS cable, but still above average. I’d say it’s roughly on par with Zowie cables. Thickness is standard, so it should fit into any mouse bungee (tested with a Zowie Camade). Cable length is 1.80 m (measured). There is a ferrite bead attached to the cable which helps lessen noise.



Build quality
Despite being a very lightweight construction the XM1 has a sturdy and solid build. There is no rattle, no loose buttons, no creaking and no flexing. The main buttons have very low pre- and post-travel, giving them a nice and snappy, fairly light click. The right click has a hollow and subdued sound on my copy, however, which isn’t present on the left one. This might be fixable by loosening the screw directly below the affected click, but I haven’t tried it so far. Side buttons have low pre- and moderate post-travel, with the forward button being the better one. The scroll wheel has very distinct and tactile steps while being very quiet to scroll in both directions. Actuating the scroll wheel requires an above average amount of force.
The XM1 comes with a black matte coating, which provides decent grip while not attracting dirt too much. I’d say it’s roughly comparable to current Zowie coatings (slightly grippier and not as much of a dirt magnet though).
On my copy there is a faint high-pitch noise (‘coil whine’) audible when the sensor is having contact with a surface. It can only be heard at certain angles and it’s not as loud as the high-pitch noise on the early batches of the EC-B series, but noticeable still. The noise may be inaudible entirely for those who are not susceptible to hearing high frequency noises in the first place. The factory has already been informed of my findings, so we may see this addressed in the future, provided that it’s not a one-off occurence.

Testing
CPI accuracy:
For this test I’m checking whether the actual CPI values match the nominally stated CPI values. The default (out of the box) nominal CPI values are: 400/800/1600/3200. The actual (tested) values, however, are: 410/816/1638/3270. In short, actual CPI is consistently higher than nominal CPI, but only by a small margin. Note that this variance will itself vary from unit to unit, although the general trend should persist.

Sensor smoothing (motion delay):
In this test I’m gauging whether the sensor performs how it’s supposed to. ‘Sensor smoothing’ describes an averaging of motion data across several capture frames in order to reduce jitter at higher CPI values, which increases motion delay. The goal here is to have as little smoothing as possible. According to specification the 3389 has no visible smoothing up to 1800 CPI and 32 frames of smoothing at and above 1900 CPI, which doubles twice at 6000 CPI and at 11300 CPI. Let’s see how the 3389 in the XM1 fares in this test.

First I’m doing a basic test in which I’m looking at a plot of the raw motion data, aggregated as xCount. In an xCount plot smoothing can be recognised by any ‘kinks’ being visible, which show up at framerate transition points. As you can see no kinks can be observed, which means that no smoothing is present.






In order to ensure that my result from the xCount test is correct I’m taking a look at a plot that compares the XM1 to a G403 (3366, no smoothing across the whole CPI range) at 3200 CPI, aggregated as xSum. Any CPI deviations between the two mice have been normalised. As you can see there is about 0.5 ms between the mice, which is within margin of error (G403 was also physically moved before the XM1) and confirms the results from above. The XM1 does not have any visible smoothing at the tested CPI steps, which is a first for a 3389 mouse.



Paint test:
This test is used to indicate any potential issues with angle snapping (non-native straightening of linear motion) and jitter. There may be some added jitter at 3200 CPI (would be expected due to the lack of smoothing), but it’s very minor. Other than that no issues can be observed.



PCS:
‘PCS’ is short for ‘perfect control speed’. The PCS indicates the maximum speed up to which the sensor functions flawlessly (i.e. without malfunctioning in some way). The nominal PCS of the 3389 is 400 IPS, which translates to 10.16 m/s. I only managed to hit a measly 4 m/s. As you can see there are no signs of the sensor malfunctioning at that speed.



Click delay:
One of the main selling points of the XM1 is its unique (and patented) analogue switch contact technology. Since this is something that hasn’t been done before, I figured some technical details would be interesting to some, which is why I’ve asked one of the developers behind the XM1 to provide a rough explanation. I’ll provide his quote in full:

“Analog Key Sensing Technology does not debounce at all. Debounce is something from the digital world. The switch itself however is still the same switch, causing the same bouncing as before, but the Analog Key Sensing takes sample snapshots at an incredibly high speed (very much like a scope) and the bouncing of the switch is summed into this analog sampling signal and therefore the bouncing is transformed into a mathematical function rather than a digital bounce (0/1), thus eliminating the need to debounce altogether. This works similar to how different brightness levels are achieved for LEDs traditionally. LEDs are pulsed to achieve that, being always driven at their max brightness for small periods of time. The more times that happens the brighter the LED is perceived. Analog Key Sensing is the reverse of this and using true analog to actually see the ‘brightness’, i.e. the key level.”

Onto the testing. I’m only able to test the click delay with a program called Bloody KeyResponse, which not only accounts for any differences in terms of hardware debounce delay but also in terms of click actuation weight, which means that a mouse with lighter clicks may return lower delay values, despite not being actually faster. In order to minimise any inconsistencies I’ve tested the XM1 against several different mice with varying click actuation weight (with the G403 having the lowest). All lenses were covered during the test, as some mice have higher debounce during lift-off. Here are the results (values in brackets indicate the delay relative to the Steelseries Ikari, which acts as a baseline):

Logitech G403 (4.2 ms): XM1 was slower by 2-4 ms
Nixeus Revel (6.2 ms): XM1 was faster by 5-7 ms
Microsoft WMO (12 ms): XM1 was faster by 14-16 ms

Based on these values (the G403 appears to be an outlier, due to the very low click actuation weight) I’d say the XM1 should be around +0-2 ms relative to the Ikari, which is an excellent result. Of course, I’m unable to test whether the ‘no double clicks’ claim holds, only time will tell that.

Software
Since the XM1 already allows for CPI level and polling rate adjustment without any software, the software is kept very basic. It allows for button remapping (in theory, appears to be nonfunctional currently), CPI adjustment (four levels, in increments of 50), LOD adjustment (2mm or 3mm, the former being the default setting) and enabling or disabling ‘ripple control’.
The most interesting option is actually ‘ripple control’. As seen in the sensor testing section, the XM1 does not have any smoothing at the default CPI levels. In fact, I went ahead and tested 16000 CPI and still could not detect any smoothing (resulting in jitter and ripple being out of control at that setting). Upon enabling ‘ripple control’, however, the usual behaviour from a 3389 can be seen: 32 frames of smoothing at and above 1900 CPI, which doubles twice at 6000 CPI and at 11300 CPI. Personally I’m a big fan of giving users control over options like this one, as both people who want high CPI but don’t like smoothing as well as people who want high CPI but don’t want ripple and jitter can be satisfied.

The Verdict
Once again I’m getting to review the first mouse release of a new company, Endgame Gear, and once again I’m impressed. The XM1 combines a very good (and unique) shape, a flawlessly implemented top tier sensor, innovative technology (analogue debouncing) and very low weight (without having to resort to holes), without having any of the rookie issues you might expect from a new competitor. My copy had some build quality concerns (button uniformity, high-pitch noise), but they weren’t major enough to negatively impact my experience with the XM1. Recommended.


Picture album
Plot album
 

·
Registered
Joined
·
1,181 Posts
skatez arent rounded :(
3rd Party skatez will be have Rounded Edges ... ( corepad will make some )
MWheel feels great ( ALPS ) , just the Middle Click is a lil bit too hard ...
 

·
Registered
Joined
·
1,809 Posts
and lastly a patented analogue switch contact technology
What's the patent number? And why would they need a patented technique in the first place? If they're using using both switch contacts, that's not something they can patent. If they're not using both switch contacts, that's inferior.
 

·
Mouse addict
Joined
·
795 Posts
Discussion Starter · #5 ·
What's the patent number? And why would they need a patented technique in the first place? If they're using using both switch contacts, that's not something they can patent. If they're not using both switch contacts, that's inferior.
I don't know either, hopefully the explanation from their engineer can shed some light on this.
 

·
lololol
Joined
·
4,380 Posts
hmm... something around 6500 max framerate

the dots falling on 3 lines (instead of 2) in the mousetester plots indicates that the timing of reading motion data from the sensor isn't stable.
 

·
Registered
Joined
·
690 Posts
"the XM1 is actually most similar to the Kana"

naaajs

It gave kana vibes on first sight, I hope the feel is similar too.
 

·
Banned
Joined
·
8,804 Posts
Great review, and, apparently, great mouse as well! Lack of smoothing is a really nice surprise.
 

·
Mouse Junkie
Joined
·
1,445 Posts
Fantastic work as usual mate!
 

·
Mouse addict
Joined
·
795 Posts
Discussion Starter · #10 ·
I've added the technical explanation regarding the advertised '<1 ms Analog Technology' from one of the developers to the review.
 

·
Registered
Joined
·
1,809 Posts
I've added the technical explanation regarding the advertised '<1 ms Analog Technology' from one of the developers to the review.
That explanation doesn't really impress me, at best it's like going from an ordinary incandescent bulb to a halogen, when you actually want a LED.
 

·
lololol
Joined
·
4,380 Posts
That explanation doesn't really impress me, at best it's like going from an ordinary incandescent bulb to a halogen, when you actually want a LED.
yup...

it's more like you already have an LED and then you stick in a low pass filter...

wall of text i posted on leddit:
if you look at the pcb it's just an RC low pass filter on each switch. and i assume it uses the mcu's own adc to read an analog level.

0. yes this is debouncing, but done with an analog filter. it is possible, though inefficient, to exactly emulate the behavior in firmware via a moving average of digital reads of the pin.

they seem to imply that the signal from a switch is analog and spends a lot of time floating between say 0 and 1. this is not true. if you look with an oscilloscope you can see that the amount of time a switch spends between 0 and 1 for a single transition is less than a microsecond (because the switch capacitance is very low).

1. this analog filtering necessarily introduces a delay. it might be small, but it is a delay nonetheless. define a signal of 1 as unpressed and 0 as pressed. for a perfect switch that doesn't mouse, the delay is equal to -tau * log(X) where tau = R*C is the time constant of the lowpass filter and 0 < X < 1 is their threshold to count the analog signal as a press. If the switch is bouncy, the delay becomes some longer by some fraction of how long the switch bounces.

Most likely they are using a somewhat large tau for robustness in debouncing, but an X close to 1 to minimize the introduced delay.

likewise the delay on release is equal to -tau * log(1 - Y). i don't know what they actually do, but what they should do is select a large Y close to 1 such that pre-releasing doesn't happen (i.e. when the switch is physically depressed but electrically loses contact momentarily. this happens as pressure is reduced on the switch). this would mean that there is more delay on releases then presses.

2. one good digital way to debounce is to not introduce any delay on presses, and a fixed amount of delay for releases. this is always what logitech does (except when the sensor is lifted, in which case they add a few ms delay for clicks, presumably to prevent slam clicks).

3. there is also the sr-latch method, which has no delay for presses. for releases technically the delay (relative to methods using only one pin) is how long it takes for the moving contact to go between the NO (normally open) and NC (normally closed) contacts. i dont remember exactly but i think this is around 1-2ms.

4. failure modes:

the resilience to bouncing for both the analog method (1) and the digital method (2) is related to how much delay is introduced on the release, if we assume both have negligible delay on clicks (i am not sure is the case for the xm1). you might be able to design specific signals where the analog method is superior or where the digital method is superior, but in general i expect that they are similarly resilient if the delay is comparable. for a sufficiently worn switch, both methods will bounce or prerelease.

for all methods, if the NO contact is sufficiently worn, there is the possibility of a press not registering. usually one will feel that additional pressure is required beyond physical actuation to achieve a click. for methods (1) and (2) there is the risk of bouncing if the NO contact is worn. for method (3, latch) there is no risk of bouncing but there is the risk of a release being delayed or not registering if the NC contact is worn. i have never heard of anyone reporting this in practice however, but this partly because very few mice use the latch method.

5. i recall that one of their reps mentioned taking bouncy switches in another mouse and claiming that the analog filtering fixes the issue if they try the same switches in the xm1. this is not necessarily a fair comparison because the process of de-soldering and soldering can affect whether a switch bounces.

tldr: mouse are digital devices. using analog filters and analog reads works, i suppose, but really has no advantages over traditional methods with 0 press delay. so i don't see the point except for marketing to people who don't know better.



edit: another thing:

it is somewhat difficult to test how much delay the xm1 has. if you just wire the pins in the xm1 to another mouse, the signal in the other mouse will be adversely affected by the xm1's low pass filter, resulting in additional delay for the other mouse. so button mashing might be easiest thing to do
 

·
Registered
Joined
·
113 Posts
Cool looking mouse. Wanna see what Razer has to offer with their new mouse before picking this up. Sounds like this is the kinzu we've been waiting for?
 

·
Mouse addict
Joined
·
795 Posts
Discussion Starter · #16 ·
How would you compare the size to Revel? Does it feel any smaller?
In hand it definitely feels smaller for me. The difference in length is the most noticeable one.
 

·
Not a Linux Lobbyist
Joined
·
1,216 Posts
Where/ when did the 3389 smoothing get tested? I'm left at q's da elite review where there are no specific numbers:

-several ms of smoothing at and above 1900dpi (will determine exact amount later)
-smoothing doubles at or above 6000dpi
-smoothing doubles again at or above 11300dpi

Did we get actual values somewhere later in time?
32 - 64- 128 frames seems... bleargh.. brupfhhh.. disgusting?

And lol the 336x framerates.. such an improvement over the 3090 huh? Like really just get a 3090 and read the registers twice a poll but no.. to sell you need 16k cpi, and some streamer to hype you up.. and reddit.
Btw has there ever been a timing-stable implementation of the 3389 lol? Razer's original was messed up too.
And why are all the new mice whining? It's like the industry's goal is to create new ridiculous issues.

As long as people buy this stuff they'll keep making it. And since gaming is now a fad, many people who can't understand technology will be in the game and support this type of product.
What a wonderful time to live in for gamers huh? I've been playing videogames since 1993 and I'm actually quitting, nothing is fun or satisfying anymore.
There is no market interest into making products like that, weather it's hardware or software.
 

·
Banned
Joined
·
8,804 Posts

·
Mouse addict
Joined
·
795 Posts
Discussion Starter · #19 ·
Where/ when did the 3389 smoothing get tested? I'm left at q's da elite review where there are no specific numbers:

-several ms of smoothing at and above 1900dpi (will determine exact amount later)
-smoothing doubles at or above 6000dpi
-smoothing doubles again at or above 11300dpi

Did we get actual values somewhere later in time?
32 - 64- 128 frames seems... bleargh.. brupfhhh.. disgusting?

And lol the 336x framerates.. such an improvement over the 3090 huh? Like really just get a 3090 and read the registers twice a poll but no.. to sell you need 16k cpi, and some streamer to hype you up.. and reddit.
Btw has there ever been a timing-table implementation of the 3389 lol? Razer's original was messed up too.
I've later compared a 3360 and a 3389 at 3200 CPI (xSum) and there was no difference in terms of motion delay, so it can be inferred that the amount of smoothing in ms is the same, and due to the framerates being very similar at low speeds the number of frames of smoothing being the same as well. At max CPI the 3389 indeed reaches ridiculous levels of smoothing.
 

·
Not a Linux Lobbyist
Joined
·
1,216 Posts
sadly flaming people with higher standards than him, and refusing to discuss or understand market criticism
Suggest one recent game that is not just a badly thought out cash grab. I'm listening.

I've later compared a 3360 and a 3389 at 3200 CPI (xSum) and there was no difference in terms of motion delay, so it can be inferred that the amount of smoothing in ms is the same, and due to the framerates being very similar at low speeds the number of frames of smoothing being the same as well. At max CPI the 3389 indeed reaches ridiculous levels of smoothing.
Yet we had to bare with pages and pages of people swearing it's more responsive than the 3366 for like how much time?
I like how saying to somebody that he is not able to understand something is considered offensive while wasting people's time and energy isn't.
 
1 - 20 of 59 Posts
Top