nForce4 NIC replacements

[Fanon]

Background:

My wife was two-boxing in WoW. Our main computer is a Opteron 165 running on a DFI LanParty NF4 Ultra-D mobo. We're currently using the nForce networking connection. The second box she uses is a Intel MacMini, with a Marvell Yukon. On average, the Yukon's latency is 50ms to 100ms faster than the nForce 4 interface.

I've updated the BIOS and loaded the latest nForce drivers. This helped, but not a great deal. I'm currently looking into this card: http://www.newegg.com/Product/Product.asp?Item=N82E16833106121 to use instead of the nForce. I've been out of the loop as far as NICs are concerned, and would like some input. Any thoughts are welcome, as well as suggestions on other solutions.

 

[Pabster]

Just out of curiousity, are you running the nForce 4 Firewall crap?

 

[Fanon]

No, I'm not. I knew it caused many problems, so I didn't bother installing it.

 

[Fanon]

double post

 

[Madwand1]

Some of those boards have dual NICs, and one of them is often a Marvell bridged via PCI; whichever port you're currently using, you could try the other port.

You should also try the advanced NIC properties; notably optimization "CPU" vs "throughput" for nForce -- try both options. Next try interrupt moderation if it's available (probably on the Marvell, not on the nVIDIA, which wraps this via the "Optimize For" setting).

50 - 100 ms seems like a huge difference; I wouldn't expect such a large difference just due to NICs.

Finally, if you really must try an add-on, I'd recommend a PCIe add-on, e.g. a SysKonnect (which use Marvell chips). There's also a Koutech on the egg with a Marvell chip, but I haven't tried it myself so couldn't comment.

 

[cmetz]

Fanon, the card you linked would be a good choice.

 

[Madwand1]

I'd suggest the Intel 1000 PT (PCIe) instead. Again, assuming that the NIC is the problem here.

 

[JackMDS]

Does any one know that the average Simple Reaction Time of human beings is 250ms?

Standard reaction time in cognitive research is measured by putting the subject in front of a Panel that has on it a Bulb and placing is the index finger of the dominant hand on a very fast, short trip, micro switch.

There is a Warning buzz about 3 sec. before the light comes on.

A timer measures the interval between the onset of the light and the closing contacts of the key.

The equipment is using components that introduce a variance of 10msec.

What does it means?

It means that it takes the brain 250msec. to process this most simple information (light On), and react upon it in aa very simple manner (pressing a button).

Take into consideration that every thing else (including computer games) is more complicated than Simple Reaction Time, and thus takes longer to close the loop of perceiving a stimuli, making a decision, and act.

In addition, Alcohol, Drugs, Food, and general body condition. and many more variables can affect the system, and make it even slower.

So what does this means?

It means many Computer Hardware considerations that people (who simply might be not such a good players) are concerned about are really just "Nonesesbse.

The Standard Varience Errors of the human neurological functions are larger than the bickering over saving 50ms (as an example) in computer?s functions.

:sun:

P.S. Human Central Nervous system uses a conduit to electrical activities substances called neurotransmitters. Chemical processes are by nature much slower than pure electron exchange as used by computers.

 

[Captante]

The 250ms stat is interesting, but aside from testing in a lab I'm not sure its relevent here.. two otherwise equally skilled gamers on two identical PC's going head-to-head, with the only difference being one's network connection is taking +/- 100ms's longer to send & receive data will result in the player with the slower connection getting destroyed every time, especially on a LAN... this line of reasoning reminds me of the rational that 60fps is the same as 160fps because the human eye can't see the difference.

 

[Fanon]

I set the optimization to throughput after I installed the drivers. I will try the other interface and report back. I didn't know it was a Marvell; I thought they were still 3Com (shows you how much I've kept up with nforce).

 

[skyking]

According to this description, you have a couple of choices.
http://www.techpowerup.com/reviews/DFI/LPNF4UD

 

[Cerb]

That reaction time bit is like not seeing more than 30FPS: it does not take into account all important variables. Not the least of which being that no client side prediction can make up for real short ping times, and in the bets case, added ping time means that much added to reaction time, but really has a worse effect. Also, much of gaming id reflex and prediction, making it even more complicated.

As noted, it looks like you already have another NIC. If not, IIRC, Intel has it when it comes to speed and low CPU usage.

 

[Fanon]

Update:

I switched over to the Marvel interface and saw no change; so I'm going to look into PCIe cards and see what available. Thanks all for your thoughts and suggestions!

 

[Madwand1]

50-100 ms is still pretty big when you consider that you can get < 10ms local pings over typical wireless for 1400 byte packets.

After trying two different NICs and some of their optimization parameters, I'd probably look elsewhere first before buying another NIC.

Maybe the NICs/drivers are having a secondary effect through CPU utilization. For that, you could try the "Optimize for CPU" option. I actually get better throughput with nForce 430 in some cases with that option.

Beyond that.. Do you have the same config on both sides? Disabled firewalls, virus scanners, other apps? Tried the computers in isolation instead of at the same time competing for limited Internet resources? Perhaps the Mac version is just optimized a bit better in this case?

 

[spidey07]

your IP stack might be out of wack.

search MS for "repair TCP/IP stack"

 

[nipplefish]

Originally posted by: spidey07
your IP stack might be out of wack.

search MS for "repair TCP/IP stack"

What he said...

I used to have that board. I never had any lag problems with either interface. The marvell seemed slightly faster than the nvidia but for everyday use either one was fine.