KT800 or Nvidia nforce3 150/250



any one know what the big differences are? i always liked the nforce boards…got a nforce2 myself now…but if i get a athlon 64 3400 chip (well im getting a free one) then i need to choose a motherboard.

i like abit but they only carry the kt800 boards
chaintech and others do make nforce 3 (150 and 250 boards)

anyone have experience in these chipsets? and whats the difference between the nforce 3 150 and 250 is it just the extra raid ports?

and no i cant get the 939 chipset boards as the chip is socket 754.

edit: ok the 150 is a 150 gigabite ethernet card where the 250 is a 250 gigabite one. man 250 gigabite ehternet thats fast…


nForce3 150 is crap and 250 is “decent” depending on how you look at it.
nForce3 has (very) poor support in other operating systems such as Linux and nVidia will most likely not do anything to resolve that issue (just look at nForce2). Feature-wise nForce3 doesn’t really have something that really blows the other cards away. I’d go for a SiS755-board such as Foxconn’s (755A01-6EKRS) which has proven to be very stable and reliable while maintaining good performance (Use google to find reviews). Regarding VIA K8T800/K8T800PRO they’re also decent chipsets although some mainboards are pretty bad while others are very good. The performance between K8T800/K8T800PRO is the same (~ +/-1%) and both SiS and nVidia are a bit faster but it’s negligible during normal use.


ack its a 939 chip 3500 so looks like ill need to go with a asus kv8 board or some other like that.


NForce 3 250 has not only dual channel memory support (only on Socket 939, 754 is single channel) but the really important thing is that the 250 has the fastest HyperTransport bus currently available. It also has a hardware firewall, and support for a lot of stuff. MSI has actually made two Socket 754 NForce3 250 mainboards, the K8N Neo Platinum, and the K8N Neo FSR. If I bought an Athlon 64 rig, I wouldn’t hesitate to go with an Nforce3 board, and personally, I’ve had good results with MSI mainboards. The K8N series is currently at the top of most reviewers’ lists as well.



Huh? K8T800Pro supports 1 GHz HyperTransport if that’s what you mean so you’re a bit inaccurate and the memory bus is (if I’m not mistaken) integrated into the CPU so Dual Channel Memory support has nothing to do with the chipset. From what I read the hardware firewall is worse compared to a software one and why would you like to have a “hardware firewall” in the first place. I would also consider nVidia’s skills in making drivers, just look the the IDE-driver for nForce2 and support for other operating system. It’s not an obvious choice.


well i havent seen many nforce3 250 939 socket boards out there best price seems to be the k8t800pros…I thought my 3200 barton supported dual channel memory or is this different than twin channel? my nforce2 board is running 2 512meg. twin (dual) channel no problem? are you saying the athlon 64 is a step down from the bartons because it doesnt support dual channel memory ? is the fx the only one that does?


Dual channel memory is over rated. Try running some RAM benchmarks in single channel and dual channel and you’ll see there’s not much difference. High FSB and clock speeds are what makes a system fast.


so i could save money is just buying a single 1 gig. ddr instead of more for dual 512s…cool.


It really depends on how you are running the board, and the board itself. Many overclockers report better stability in single channel mode. But again it depends on the board. I would test a board both ways before making a decision, and as always, don’t buy cheap RAM.


To start off with, the hardware firewall means that the job is offloaded from the CPU. I have only heard good things about this from reading review after review. Why would I like a hardware firewall? Because it’s on the moment the system is on. Software firewalls like Windows XP’s, or ZoneAlarm, Sygate, BlackIce Defender or others, load during system startup, and there are points when the system is temporarily vulnerable, as network drivers load prior to them. Full-time protection certainly appeals to me.
As for the IDE drivers, NVidia has some work to do on them, but I wouldn’t take this as an example of every driver of every NVidia piece of hardware to be bad, as the other chipset drivers work fine, as has the vast majority of NVidia video card drivers. I’ve seen enough instances where VIA had issues with their drivers that I’d say this could go both ways. A number of current VIA mainboards also have trouble with their hardware PCI/AGP lock, a problem which VIA has supposedly fixed, but those mainboards are out in the channel, and it could be quite difficult to figure out which manufacturer or which model of board has a working lock if you’re looking to play with overclocking your processor.

I don’t have an issue with VIA, btw, I’ve had a number of their boards and been satisfied with them. I’ve just had even better performance and features with the Nforce chipsets, and equal or better stability.


Well, the A64’s and the old Socket A CPUs works a bit differently but to make it easy to understand you can say that the A64 have the memory controller builtin which greatly gives memory bandwidth a push compared to Socket A (where CPU first have to access the chipset). The CPUs that officially supports dual memory channels are 939- and 940-socketed ones. Rdgrimes is right regarding the performance enhancement of dual channel except in one scenario, nForce2 IGP (nVidia’s nForce2 based integrated graphics solution). You can see a ~15% (About that if I’m not mistaken) performance boost but that only applies if you use the integrated graphics otherwise it’s “useless”. So yes, you can save money by buying one memory stick. (Actually it’s more expensive to buy a single stick than two 512, at least where I live)
A general recommendation is to buy branded ram by a manufactorer that’s commonly known, Corsair, Kingston, Samsung, Crucial etc. Be a bit careful when choosing ram though, some chipsets doesn’t like CAS 3.

@ LoneWolf
Can you point me to a review where they actually tests the firewalls performance? I’m very doubtful it’s all that great (especially looking at the one for nForce2 which performs like shit (my P166 probably performs better)). Regarding the hardware firewall I’m not so sure it’s always on when actived, it’s probably turned on when the driver loads and that time and loading a system service (software firewall) is about none. You also have a lot more flexibility (if needed) using a software firewall since the nVidia one is only a personal one. Keeping that in mind you’re also pretty screwed if you want to try Linux, *BSD. Sure, they provide source code but no papers and no native support which most likely will lead to a dead end in the long run. The PCI/AGP has been fixed for quite some time and if you’re unsure you can just google for a review.


The NForce3 150:

  • 1200 HyperTransport Bus
    The NForce3 250:
  • 1600 HyperTransport Bus
    There will be no NForce3 chipset for socket 939. The NForce4 chipset, due to arrive before long, will. On paper, the NForce4 looks to be an awesome chipset.
    However, in any case, I am nowhere near sold on the NForce socket 754 chipset.
    I do not recommend it to any customer who’s in the marktet for an Athlon64 board.

If you’re looking into a socket 754 Athlon64 motherboard…look to the VIA or SIS offerings. I will put these 2 chipsets up against the NForce3 any day.

I currently run the SIS755 chipset on the Foxconn motherboard and couldn’t be happier.



The NForce3 Ultra is already available for S939 I believe. The MSI K8N Neo2 Platinum is a board based on this chipset:


BTW Diizzy, I don’t have complete tests on the firewall listed, but The Tech Report has a good article on the chipset. One note they have at the beginning of the article goes as follows:

[i]* A hardware-accelerated firewall — I’ll look at the nForce3 250Gb’s firewall in more detail in a moment. However, it’s worth noting here that the firewall in the only native, hardware-accelerated one of its kind. Acceleration should help reduce the firewall’s CPU utilization, and NVIDIA is even letting third-party software developers take advantage of the nForce3 250Gb’s dedicated firewall hardware.

The firewall’s hardware components have benefits beyond lower CPU utilization, though. Because it partially resides in hardware, the firewall is enabled instantly when a machine is powered on. Software firewalls that rely on drivers and other programs can only protect a machine after it boots into the OS and all the necessary drivers and programs are loaded, leaving a small window for attacks to take hold.[/i]

The article can be found here:


My bad…forgot about the NEO2.
Crap…we just got that board in last week.
I apologize for being an idiot.

In any case…I’d wait for NForce4 if you want an NVIDIA A64 chipset.


New Socket 939 mobo roundup at Anandtech. I don’t take every hardware review they do as gospel (esp. optical drives, where I find them lacking in experience) but their motherboard reviews are pretty good and should highlight the VIA KT800Pro and NForce3 Ultra performance. MSI and Gigabyte’s NForce 3 Ultra boards look particularly impressive, if you’re a VIA fan the MSI, Abit, and ASUS boards look good as well.



Is the nforce 3 really as bad as everyone is making out…i’m probably going to go for that chipset for my A64 laptop…is it a bad idea?



@ LoneWolf15
To be honest with you I’m very sceptic about this so called “hardware firewall”. If it’s all that great why haven’t anyone done it earlier? I’ve yet to see tests on it too.


After being a hardware tech for nearly ten years, I’ve seen plenty of hardware debates over hardware vs. software solutions, Dizzy. And just because something wasn’t done earlier doesn’t mean it’s something to be skeptical of; in 1890, horseless carriages “hadn’t been done earlier”. Technology advances, and new products appear. Memory technology for example, has increased leaps and bounds in the past decade, and so have CPU’s. Why weren’t they done earlier? At least in part, because the fab technology wasn’t available to create such a complex design on such a small surface area, and software wasn’t complex enough to drive demand. In the case of firewalls, up until the late 90’s, broadband internet was a rarity at home, and people didn’t need security the way they need it now.

Truth be told, people have done hardware firewalls earlier, just not in the way NVidia does. The Cisco PIX for example, is just one version of a hardware firewall, and there are others, though up until now, they have been restricted to enterprise due to being cost prohibitive for a small home network. Plenty of people build a box running Linux just to serve as a router/firewall/etc. . If one built this same self-contained OS and firewall into a chip, or an intelligent box (and of course, this has already been done), what would there be to be skeptical of, as they have already been proven? Firewalls after all, are no longer rocket science, used mainly as intelligent port blockers under user control, more advanced firewalls also being able to detect types of malformed packets and specific types of web attacks.

Since not all of us have dedicated hardware to make into a firewall (or limitless funds to pay for the electrical bills that the additional hardware provides), a hardware firewall built onto a mainboard is a pretty doggone good idea, IMO. Done properly, it can be OS independent, either administered through having a remote web interface, or through other configuration means (possibilities including through BIOS setup, or multiplatform software, such as a Java applet). This part only being for the administration of course, which is how many enterprise web appliances are today (web content filters, proxy servers, and the like), as the firewall would operate independently of them once configured. Since the firewall is not dependent on the OS, it is on regardless of whether the OS is fully loaded or not. Those of us who have had to cure several hundred machines of Sasser, MSBlast.32, or the Nachi worm know how big a deal this is, as these worms can strike the moment they find an open machine, when a software firewall may not have fully loaded up (note: Windows XP Service Pack 2 is supposed to improve upon this flaw, but I’ll wait to see how well it works in practice). Because the firewall is in hardware, it also will not take up CPU time or memory resources, another big plus. It is less likely that a hacker could design a trojan to defeat it as well, something that happened recently to ISS’s BlackIce Defender when a former disgruntled employee coded a program to kill it (patches were subsequently released).

I haven’t seen a whole lot of testing on software firewall solutions, let alone NVidia’s hardware firewall, however you can bet any vulnerability would be quite a bit of bad press for NVidia. No matter how good any firewall is, there is no perfect firewall. But, I’d rather have one implemented in hardware for all the reasons listed above.

P.S. None of this is a knock on you Dizzy…I hope if anything, you consider this a debate, not an argument.


whats the difference between a hardware firewall on the motherboard and one like i have on my cable router?


Aside implementation and features, they are both still hardware-driven firewalls. One main distinguishment, as far as pro’s/con’s go is that software driven firewalls are updated via firmware, and are limited somewhat because of limitaion of space and other variables. Software-driven firewalls are obviously easier to update and expand on because you have the size of your harddisk and you can change it as you please.

Generally, the concensus is, Hardware-driven firewalls have been thought of as harder to bypass.

***The best solution, is to use BOTH in conjunction.