ABOUT BENCH FORUMS PODCAST LOGIN REGISTER
PC Components
Smartphones & tablets
Systems
ENTERPRISE & IT
GUIDES

nForce4 SLI Roundup: Painful and Rewarding

 by Wesley Fink on February 28, 2005 7:00 AM EST
108 Comments | Add A Comment
108 Comments

Ethernet Performance

We have often talked about the throughput advantage of PCI Express compared to PCI and its impact on Gigabit LAN performance. To show you this, the new motherboard test suite needed to include LAN performance measurements.

The Windows 2000 Driver Development Kit (DDK) includes a useful LAN testing utility called NTttcp. We used the NTttcp tool to test Ethernet throughput and the CPU utilization of the various Ethernet Controllers used on the nForce4 SLI motherboards.

We set up one machine as the server; in this case, an Intel box with an Intel CSA Gigabit LAN connection. Intel CSA has a reputation for providing fast throughput and this seemed a reasonable choice to serve our Gigabit LAN clients. At the server side, we used the following Command Line as suggested by the VIA whitepaper on LAN testing:
Ntttcps -m 4,0,‹client IP› -a 4 -l 256000 -n 30000
On the client side (the motherboard under test), we used the following Command Line:
Ntttcpr -m 4,0,‹server IP› -a 4 -l 256000 -n 30000
At the conclusion of the test, we captured the throughput and CPU utilization figures from the client screen.

Ethernet Throughput

Ethernet Overhead

While Marvel PCI has a bit lower overhead, the throughput for Gigabit LAN on PCI is almost 30% lower than Gigabit on PCI Express. The nVidia on-chip PCIe LAN also exhibits lower CPU utilization than the full Gigabit LAN on PCIe. It has also been reported that enabling nVidia Active Armor lowers CPU overhead even further.

All Ethernet tests were performed with standard frames, but Gigabit Ethernet supports Jumbo frames as well. Jumbo frames will theoretically provide a further reduction in CPU overhead. We have seen test results that show the combination of Active Armor and Jumbo Frames, which can reduce CPU utilization below 10%, a very respectable performance for on-chip gigabit LAN.

Firewire and USB Performance Audio Performance
PRINT THIS ARTICLE
Comments Locked

108 Comments

View All Comments

  • fitten - Tuesday, March 1, 2005 - link

    quote:I still do not understand why this argument is so popular. Why is the general assumption that purchasers of SLI capable boards will immediately want to jump into a dual-card config? The idea is flexibility. Sure, 2 6800's are expensive now, but they will inevitably get cheaper.


    Well, if history serves as a measure... by the time that 2nd board becomes cheap enough to justify its cost, there will be a new board out (say, the nVidia 7800) that will be as fast, or faster than, the SLI combo.

    I used to buy motherboards with two sockets for this very reason (flexibility to upgrade to two CPUs later) until about twice doing this I learned that by the time I was ready for that 2nd CPU, there was one out that was faster than both put together.

    Computers change too fast. If you perpetually buy on the bleeding edge, you cannot plan any upgrades past ~6 months and definitely not past 12 months. By that time, you'll throw away what you have and get the NextBestThing(tm). Buying SLI is bleeding edge. Saying that you'll buy the upgrade card in a year is just a rationalization to buy the bleeding edge now.

  • Aquila76 - Tuesday, March 1, 2005 - link

    Yeah, that's right. Some apps run slower with SLI because nVidia hasn't SLI optimized the driver for that app (so it can then only utilize one card) and the SLI setup uses some overhead, resulting in slower results. Any new game/benchmark will use SLI just fine. The results in Half-Life and Doom 3 as well as if you add the config for stuff like NFS:U2 and whatever are well above one card though.

  • Sunbird - Tuesday, March 1, 2005 - link

    Is my brain screwed up or are the 3Dmark03 single scores higher than the SLI scores???

  • chup - Tuesday, March 1, 2005 - link

    too bad, i thought the MSI was the one to get after nforce2.

  • sphinx - Tuesday, March 1, 2005 - link

    From this review, I have come to the conclusion that ASUS is slipping. I have always been a supporter of ASUS but, I think this review shows how much ASUS is all about the money and not making quality products. Right now I am waiting for manufacturers to get the VIA chipset working properly. I haven't seen many news or reviews on VIA's new chipset. One other thing. Who in their right mind would spend close to $250 on nVidia's NF4 if there is really no significant performance jump from the NF3.

  • bigbusa - Tuesday, March 1, 2005 - link

    You mentioned the asus manual says use a 500+W PS. if you read the Asus users guid the sli 6800 ultra system also has all pci slots used, all memory dims full, 2 optical drives, and anassortment of other stuff. and they recommend a 500+W, but a 350W PS for a dual 6600GT. See below.

    500+W ps for 55FX, 2x6800 ultra, 4ddr dims, 4 HD's, 2 optical, 1 pcie 1x card, 3 pci card, 1 1394, 6 usp devices. (shit thats alot of gear)

    350W for a 3400(64bit 939), dual 6600GT, 2 DDRdims, 2 hd's, 1 optical drive, no pcie 1x, 1 pci card, no 1394, and 3 usb devices.

    SO the article is misleading a bit.

    The review also did not cover any quad displays and problems one may encounter when setting this up.

  • Reflex - Tuesday, March 1, 2005 - link

    #70 - None of these boards support ECC. The reason for that is that such support would be implemented by the memory controller, not the motherboard manufacturer. In this particuliar case the memory controller is integrated into the CPU. AMD has a line of CPU's that have ECC support, they are called the Opteron and are designed for workstations and servers.

    In the home user market ECC does not significantly impact stability but it does harm performance by a small amount which is why the feature is not generally available on consumer solutions.

  • 1955mm - Tuesday, March 1, 2005 - link

    All in all I think that this is the best review of Socket 939 SLI boards that I have seen. I particularly liked the attention paid to storage and I/O capabilities. My one criticism is that although comments were made regarding stability, and a link was made between overclocking and stability, there was no discussion of ECC support. If system reliability is discussed, ECC should not be ignored. As far as I can tell, the only board supporting ECC is the ASUS board. Over the years I have found it difficult to get accurate information on ECC support, having been given misleading information on occasion by both MSI and ASUS.

  • Aquila76 - Tuesday, March 1, 2005 - link

    D'oh, *SoundSTORM Savior*

    That's it, I'm off to bed. It's quarter of 1:00AM and I have work tomorrow. Uh, today.

  • Aquila76 - Tuesday, March 1, 2005 - link

    To everyone hoping the MSI upsamples analog 5.1 to Dolby Digital - I don't think so. Like any Creative card, it can either downmix DD-EX/DTS-ES 7.1 streams to 4/5.1 speakers (which is what page 5-11 of the manual is actually talking about doing), decode DD/DTS on-card to 5/6/7.1 speakers (via analog or 'Digital Out', Creative's proprietary digital link for their speaker sets), or can just pass the Dolby Digital/DTS 5/6/7.1 signal (now via the SPDIF coax/optical cable) to any outboard decoder.

    I say this because I have the same exact chip on a stand-alone card, and it does not upsample analog sound to Dolby Digital, like SoundStorm did. 'Digital Out' simply let's you use a proprietary Creative Digital DIN connector to connect one cable from the soundcard to the Creative speaker amp (like on a DTT3500 that I use).

    I also find it highly unlikely that Creative would license a DD Live capable chip to only one manufacturer when they have yet to produce one of their own cards with this feature.

    *Keeps waiting for a SoundStrom Saviour*
Copyright © 2024. All rights reserved.
BENCH
TOPICS
FOLLOW
ABOUT

Log in

Don't have an account? Sign up now