USB 2.0 PCIe recommendations

Everything about latency. Tips, testing methods, mouse lag, display lag, game engine lag, network lag, whole input lag chain, VSYNC OFF vs VSYNC ON, and more! Input Lag Articles on Blur Busters.
Bobo
Posts: 83
Joined: 05 Jun 2018, 11:44

USB 2.0 PCIe recommendations

Post by Bobo » 25 Oct 2020, 22:21

Can someone recommend me some USB 2.0 pcie card models?

Must definitely be USB 2.0

Ferr0
Posts: 31
Joined: 26 Jan 2021, 10:40

Re: USB 2.0 PCIe recommendations

Post by Ferr0 » 01 Feb 2021, 20:20

Kind of a late reply, but I bought both:

the Belkin
https://www.amazon.com/Belkin-Hi-Speed- ... 700&sr=8-3

and

the Startech
https://www.amazon.com/StarTech-com-Por ... 700&sr=8-4

These are the only two USB 2.0 PCIE cards I could find on amazon. I bought them because I have one USB 3.1 hub on my motherboard and I have an audio interface and a 1000hz wooting one keyboard connected to my motherboard's USB and I wanted a separate chip for my mouse. I've heard some people say not to use PCIE USB cards because they add latency but I haven't noticed any. I don't have a latency monitoring tool and I don't think this is really relevant but I achieve around the same scores on human benchmark and I've actually gotten my lowest scores with the PCIE card with some other tweaks I may also have just been really awesome for two days straight.

I tried doing research between the two cards and they do use different chips. The Belkin uses a NEC D720101GJ chip and the startech uses some VIA chip. All I was able to find regarding that is some post on a forum from 2009 saying the Renesas (NEC) D720101 is a really good chip and VIA chips come close to it. Either way I don't think it matters and I'm not even sure if the D720101GJ is the same as the D720101 chip or what the VIA chip's model number even is.

I first tried the Belkin and to me I immediately felt a difference compared to my motherboard usb, it kind of felt like disabling fullscreen optimizations at first. The best way to describe it was that my aim felt more consistent. If this effect was real and not placebo I would chalk it up to my mouse not being drowned out by my keyboard's polling rate.

My mousegraphs were more consistent on the USB cards than on my motherboard. I will say though that when I first got the cards and used the mouse tester graphs I got insanely clean graphs but when I checked my motherboard's usb 2.0 port graph it was the same. I think the main takeaway is that outside of keyboard interference the biggest effect on mousegraphs will be the driver/OS interaction. I have consistently noticed that I can eliminate spiky mousegraphs right after startup by killing dwm.exe and having it restart. Why?

Either way to me both of them felt pretty much the same. My personal favorite is the Belkin, because its blue, but I think either works. That post from 2009 is dated and there was no proof of what the guy was saying. I don't think there is really going to be a difference in latency or anything between them and the main benefit of them is to just get your mouse on a separate USB chip. If anyone actually knows any specifics about the USB chips themselves feel free to correct me. Fyi I have it installed in the last PCIE slot on a z490 aourus xtreme motherboard and it shares bandwidth with an NVME drive. I hope this wasn't too verbose, I just wanted to share my entire experience with both cards.

It might be interesting for me to test motherboard vs PCIE card with just my mouse connected on either one because last time I left my keyboard connected to my motherboard, big mistake I know, but I think I'll save that for a day when I'm really bored. Either way I don't notice any negative impact from having my mouse on one of them and if you really don't like one after buying it amazon's return policy is pretty great.

TLDR; There are only 2 USB 2.0 PCIE cards on Amazon. Buy both and try them yourself then return the one you don't like.
Last edited by Ferr0 on 03 Feb 2021, 04:00, edited 2 times in total.

User avatar
RealNC
Site Admin
Posts: 3741
Joined: 24 Dec 2013, 18:32
Contact:

Re: USB 2.0 PCIe recommendations

Post by RealNC » 02 Feb 2021, 02:40

Bobo wrote:
25 Oct 2020, 22:21
Can someone recommend me some USB 2.0 pcie card models?

Must definitely be USB 2.0
Out of curiosity, what's the use case here?
SteamGitHubStack Overflow
The views and opinions expressed in my posts are my own and do not necessarily reflect the official policy or position of Blur Busters.

forii
Posts: 218
Joined: 29 Jan 2020, 18:23

Re: USB 2.0 PCIe recommendations

Post by forii » 09 Feb 2021, 12:47

Isnt 3.0 for mouse and keyboard better?

iceboy
Posts: 22
Joined: 04 Oct 2020, 14:22

Re: USB 2.0 PCIe recommendations

Post by iceboy » 24 Jul 2021, 15:07

usb-cards-1.jpg
usb-cards-1.jpg (1.48 MiB) Viewed 6423 times
I tested both cards with Corsair K100 OPX keyboard at 4khz polling. Testing setup is 9900K 48/47, Z390, my DDR4-2400 CL9 on Win7 x64 SP1 with KB3125574. Tested for a week: Monday first NEC then VIA, Tuesday first VIA then NEC, etc. Also sometimes switch to the third card with UPD720202 or Z390 chipset ports.

The USB 2.0 cards clearly have lower latency but also worse consistency. Both cards use a PCI to PCIe bridge - the Belkin one uses Renesas D720101GJ chipset with TI XIO2000A bridge, and the Startech one uses VIA VT6212L chipset with Pericom PI7C9X111SL bridge. I suspect the bridge is making things worse because: the feeling is very sensitive to the PCIe latency timer option in BIOS, and plugging the card to the PCIe slot directly connected to the CPU (yes, it makes graphics card x8) performed much better than connecting to the chipset slots.

As of USB 3.0 options, the UPD720202 feels very consistent but with higher latency, and probably higher than the ASM1142 I've tested before. The Z390 chipset ports (with the Intel mod driver) all come with huge latency (higher than any of the cards), and there is an interesting effect: with MSI enabled, 3.1 > 3.0 > 2.0, and with MSI disabled, 2.0 > 3.0 > 3.1.

Summary:
Latency from best to worst: Belkin/NEC, Startech/VIA, ASM1142, UPD720202, various Z390
Consistency from best to worst: (can't conclude between USB3 cards but they seem to be more consistent than USB2 ones), Startech/VIA, Belkin/NEC

With all of the current options I prefer the Belkin card, but it's obviously not perfect. Looks like there is no USB 2.0 cards with native PCIe and we will have to look for USB 3.0 cards with good 2.0 support.

Bobo
Posts: 83
Joined: 05 Jun 2018, 11:44

Re: USB 2.0 PCIe recommendations

Post by Bobo » 24 Jul 2021, 17:47

have you tried using a power cable directly from the psu into the 2 usb cards that support it , and test them like that?

empleat
Posts: 149
Joined: 28 Feb 2020, 21:06

Re: USB 2.0 PCIe recommendations

Post by empleat » 24 Jul 2021, 21:21

BTW pci-e cards are not a holly grail, they usually cause huge DPC latency, just a warning!

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: USB 2.0 PCIe recommendations

Post by Chief Blur Buster » 24 Jul 2021, 22:56

empleat wrote:
24 Jul 2021, 21:21
BTW pci-e cards are not a holly grail, they usually cause huge DPC latency, just a warning!
While this is true -- PCIe may make things worse some of the time -- but not always.

For other readers:

If you're getting a PCIe card for a high-pollrate device, it's generally a good move (usually) if you've got lots of USB accessories. You want low latency jitter for a Razer Viper 8KHz. You don't want random latency effects. As long as you're using dedicated lanes to the CPU, the DPC latencies over the longer circuit lines of a PCIe lane, often tend to be more consistent when given its own uncongested PCIe lane and uncongested USB controller chip; thus outweighing DPC latency for the mouse.

So it's possible it's preferable to tolerate higher "stable" (jitter free) DPC latency in exchange for lower mouse latency jitter, if the randomized non-DPC latency is causing more problems with aimfeel. As a metaphor, consider comparing RANDOM(1-to-10) latency, versus exact 2.5 of exact perfect stable latency!? Predictable latency can sometimes be superior;

The fetishization of absolute latency neglects the random latency factors, which can also be killer too. Also, you're not necessarily moving all USB devices to PCIe -- just the mouse only or keyboard only, while keeping the other on the motherboard chipset. If you're lucky to have multiple high performance motherboard chips, you can just play port roulette until things look good latency-wise and jitter-wise (see screenshots below) -- but if you're plugging USB devices to the hilt (external disk, webcam, VR headset, etc). The onboard USB chips often leave much to be desired when you're clogging up the port... Here, the PCIe USB can be the savior here to clean up your high-pollrate.

I can even see my mouse cursor occasionally jitter if I'm sharing the USB chipset with other high-Hz devices + external USB SSD + VR headset at the same time. That's a bigger poison than a PCIe card when you've got congested ports that are simultaneously busy. And when I plug my mouse to a different host USB chipset (even if PCIe), the jitters stop. Perhaps the lag is higher, but the cursor stutter (which will also lead to mouseturn stutter in games) is the worse poison. Random USB latency so terrible, that it shows up as visible mouse jitter;

You can try vacating all your onboard ports (Razer 8KHz on motherboard, everything else on PCIe USB) but vice versa may be preferable since you have fewer events of high DPC latency since you only have 1000 to 8000 transmissions per second from a mouse over PCIe USB, instead of the millions of transmissions per second from all other PCIe lanes -- so your average DPC latency may not budge much; it's worth testing both ways. Either way, it is important to avoid USB hub congestion for your critical competition devices.

Because of all these real-world factors of clogged USB ports of many accessories, consistency improvements happen more often than not -- provided you choose an uncongested dedicated PCIe lane + one high-Hz USB device per host USB chipset. Yes, that means you may need to play PCIe slot roulette until you find the slot that doesn't contend with any existing devices. You may burn lanes (e.g. plugging a x1 PCIe USB port card in a x4 or x8 PCIe slot) just to get what you want, but so be it. And you might even be PCIe 2.0 instead of PCIe 3.0 or 4.0, but then again, that generally isn't problematic for a low bandwidth device like a mouse, as long as pollpacing improves...

There was tests that PCIe cards massively improved Razer 8KHz mice because of sidestepping the port roulette effect (see this thread)

viewtopic.php?t=7618

YMMV -- some USB chips on motherboards are superior to others -- but tolerating a very slightly DPC latency can be the lesser of evil compared to some really crappy chipset USB implementations -- so test both ways and see how the mouse feels;

Razer also did some internal PCIe-USB tests and found 8KHz Viper sometimes perform better when given a dedicated PCIe lane + dedicated USB controller. It may not always be true, but it may sidestep a congested USB controller, or shared PCIe lane (e.g. USB chip sharing with other devices on same lane) or bad chipset implementations.

Another thing to consider is PCIe lanes may go through the chipset instead of directly to the CPU; the implementations with direct-to-CPU PCIe lanes may perform better. But that congestion can be the lesser evil versus congested USB hub silicon (hub traffic congestion, USB processing congestion). It's all a pick-your-poison game; so don't avoid testing an unturned stone because of assumptions when USB cards are a mere $30 each!

I agree it is not a holy grail but I don't want to discourage people from trying such a cheap fix;
Chief Blur Buster wrote:
16 Oct 2020, 00:29
Houston, we have a problem.

Especially with 4000 Hz keyboards and 8000 Hz mice being used simultaneously.

I just helped someone solve a high-Hz USB performance interference problem between two high-Hz USB devices:

Problem: High-Hz poll performance interference between 1000 Hz keyboard + 1000 Hz mouse
MaxTendency wrote:
12 Oct 2020, 16:35
While I'm pretty convinced that 8k hz mouse polling is going to be not only noticeable but also impactful , I'm starting to wonder how much keyboard polling will affect this. High keyboard polling has known to destabilize the mouse polling and vice versa.

Image
For example this is a 1khz mouse polling on an optimized setup with a 125hz keyboard. As you can see the variance is quite little, barely 1hz.

Image
This is the same mouse but the keyboard is wooting one set to 1k hz. All of a sudden the variance is 20hz. Looks like windows can't even fully handle 1k hz keyboard and mouse at the same time.

With keyboards now supporting 4k hz polling like the Corsair K100, I'm curious how will this affect the stability of 8k hz mouse polling. Seeing that a 1k hz keyboard is enough to destabilize a mouse set to just 1k hz, a 4k hz keyboard would probably trash the stability of a 8k hz mouse.
Solution: Serparate USB controllers and PCI bus lanes per 1 high-pollrate device
MaxTendency wrote:
15 Oct 2020, 20:58
Small update, using the asmedia port for keyboard (the lowest one, right next to the usb-c port) while using top port for mouse seems to minimize if not remove the impact of high keyboard polling on the mouse polls.

Image

Blue is the mouse and red is the keyboard. This combo provided the best polling, pic of polling attached below.

Image
Yeah, that was what I thought. Glad my recommendation helped!

With ultrahigh poll rates, you really need to isolate to one USB chip (and/or separate bus lanes) per high-pollrate device.

PCI-Express USB cards are also another solution that can help this. Plug the keyboard into the motherboard USB, and plug the mouse into the PCI-Express USB. Or do do USB port roulette until you find jackpot. Keep the adjacent ports empty (port above/below a plugged-in high-Hz USB device) because they often share the same USB controller.

This would probably become a staple recommendation of the new 2020s-era "Blur Busters Mouse Guide II"

Until the motherboard manufacturers "keeps up with the Joneses" and have a dedicated-USB-chip keyboard port, and a dedicated-USB-chip mouse port for the 2020-2030s esports era of 4000 Hz keyboard + 8000 Hz mouse + >360 Hz monitors + RTX 3080+ framerates.

This Grand gaming computer upgrade supercycle is going to be very interesting. I rarely see so many concurrent upgrades happen. Those now seem to happen only once every 5-10 years, rather than every 2-3 years in the 1990s-2000s. We are seeing a major Vicious Cycle Effect tick-tock (multiple concurrent frequency upgrades).
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

empleat
Posts: 149
Joined: 28 Feb 2020, 21:06

Re: USB 2.0 PCIe recommendations

Post by empleat » 25 Jul 2021, 10:56

Chief Blur Buster wrote:
24 Jul 2021, 22:56
While this is true -- PCIe may make things worse some of the time -- but not always.
I agree that's why I said usually :lol:

SheikhMaCoke
Posts: 1
Joined: 10 Jul 2022, 11:03

Re: USB 2.0 PCIe recommendations

Post by SheikhMaCoke » 10 Jul 2022, 12:28

Chief Blur Buster wrote:
24 Jul 2021, 22:56
empleat wrote:
24 Jul 2021, 21:21
BTW pci-e cards are not a holly grail, they usually cause huge DPC latency, just a warning!
While this is true -- PCIe may make things worse some of the time -- but not always.
[...]
Hi Cheif,

I'm looking for some clarification when using the mouse tester! My aim is to use my Razer 8khz mouse with the max 8000hz polling rate without any lag/issues! Please see attached pictures of the razer mouse when using 8000 polling rate, can you confirm if the mouse tester software is the correct version & also confirm if the graphs shown are good or bad before I go and buy a PCIe USB card!

Many Thanks :D

Image
Attachments
razer 8khz mouse tester.png
razer 8khz mouse tester.png (2.36 MiB) Viewed 3976 times

Post Reply