OC effecting the network

[BTRY B 529th FA BN]

I've been pushing my W3690 lately (1366 socket overclocking) Seems pretty stable. I've ran IBT (passed,) Real Bench 30m (passed,) P95 blend (passed an hour worth) and some memtest memory stressing.. all passed. Is there a way to check and see if my packets are in good shape and being delivered, and received properly?

thanks

 

[dave_the_nerd]

It shouldn't have effected anything, unless you overclocked your NIC.

If you have a network share handy, an easy way to check is by generating a bunch of random chunks of data, copying them to the network share, copying them back, and comparing them to the originally generated data by checksumming it.

 

[frowertr]

The networking stack does this automatically if you are using the TCP protocol with error correction, checksum, ACK, etc...

But...

Wouldn't the NIC really be the one responsible for this anyway and not necessarily the CPU?

 

[frowertr]

Dave got me by a minute!! ;-)

 

[BTRY B 529th FA BN]

Well I have an Intel NIC on one of the SLI links and the QPI is overclocked which effects the stability of everything on the PCIe lanes, afaik. I've never ran my QPI speed at 4Ghz for X58 that's near the limit, from what I've been hearing. I'm thinking Real Bench would push the sub system the most and it seems ok so far.. :hmm:

 

[dave_the_nerd]

Well I have an Intel NIC on one of the SLI links and the QPI is overclocked which effects the stability of everything on the PCIe lanes, afaik. I've never ran my QPI speed at 4Ghz for X58 that's near the limit, from what I've been hearing. I'm thinking Real Bench would push the sub system the most and it seems ok so far.. :hmm:

RealBench doesn't check the network subsystem. Unless you're talking about something else.

Most "network benchmarks" are pretty superficial though. Your best bet is to script a few dozen simultaneous rsync transfers and checksum data on both ends of the transfer to verify reliability. (TCP should take care of that, but...)

You can set up a ping storm or something from another machine to see if you drop packets. (Essentially DDoS your box form another couple boxes on the same network.)

 

[BTRY B 529th FA BN]

So I understand this correctly. If the card is in a PCIe link, and the PCIe link flows data through the QPI bus both ways, you are saying there shouldn't be an effect if the QPI bus isn't stable. Correct?

 

[Railgun]

This is more a function of network testing than anything. If you're looking to test, a BERT tester is your best bet. Else, if you can run a test such as an RFC2544, that also works.

As I assume this is a home PC, if you can setup such a test within your own LAN, all the better. Else, too many other variables to test external connectivity.

I'm pretty sure such software exists that's cheap or maybe even free.

 

[dave_the_nerd]

So I understand this correctly. If the card is in a PCIe link, and the PCIe link flows data through the QPI bus both ways, you are saying there shouldn't be an effect if the QPI bus isn't stable. Correct?

TCP protocol is designed to work over unstable/intermittent connections, and does error checking and retransmission as needed. But that slows it down.

Some other protocols (UDP, ARP, etc.) don't.

But most of your network traffic is over TCP, so your network is doing its darnedest to make sure you don't a problem.

 

[BTRY B 529th FA BN]

Ok. I'm not a networking guy so all the suggestions might as well be in Chinese. Thanks for the help so far