I don't think there's any simpler proof as to why on-board chips aren't bad... how about running a full gigabit speed while running all CPU cores at 100% under Windows Vista? How much more of a challenge do you want?

On-board NIC Challenge

Orthos by default runs with LOWEST priority, meaning it'll give away CPU power at any time to any other program that might need it. I want to see a Priority 9 test.

I never said all onboard chips are bad. Onboard VIDEO still sucks though.

Bah. I don't really see a point. Onboard NICs are plenty fast as this shows (990mbits/sec), they're quite often PCI-e connected (so almost nothing to gain over that in the first place), and usually have most modern NIC features like TCP checksum offload, large send offload, network boot roms, etc. They also have every other "standard" feature you'd expect out of a NIC or hardware in general (ACPI power states, WOL, etc) CPU usage is very minimal too. Seriously, that's enough for all but the most extreme cases (and then you're usually gonna go for more than one NIC anyways -- and some boards even have dual gigabit NICs too)


Others have however.


I still 100% disagree. Non-gaming video suck at gaming, sure. No one's gonna say different. But just like me & cluberti said, onboard video is plenty for a LOT of people (non-gaming usage). I see no difference at all (performance wise) between a pair of $1000 video cards taking 100W each and onboard video. What you seem not to get is that we're not all gamers.

I used to as well. But onboard chips have evolved a lot.



You don't need a pair of $1k video cards, as a matter of fact i couldn't afford to spend more than $250 on my video card. However, a $40 card still runs circles around your onboard, uses the same amount of power and can provide useful features like HDMI output for a big display, you said you are watching HD video yourself.

How exactly? And don't say 3D performance as that's 100% irrelevant. I don't see anything wrong with it at all. In short: what makes the $40 card better at things like paying your bills and such.


I very much doubt it, any proofs of that?


Which are also found on a lot of recent onboard video cards

I think video is gonna go much like the rest of hardware in the long run (onboard), and AMD is paving the way.

how about a true full load test of a onboard nic against this:

http://www.newegg.com/Product/Product.aspx...Tpk=killer%2bm1

What about on-board sound? I have an Audigy 2 card, but I notice that the on-board sound is about as good quality-wise. Of course, the Audigy is much nicer in a lot of ways outside of the sound that it delivers, but the question I would have is whether or not most sound cards today would exceed what is on the board for sound for most applications.

Maybe that's the discussion point for this thread - when does on-board become "not enough"?

Just to satisfy those who believe that Prime95 wasn't hogging enough CPU during the network bandwidth test, I reran the test using the same conditions except this time I made sure that Prime95 was the only High Priority process (as you'll see in the screenshot) and I actually even beat my last record of 990Mbit/sec and this time topped out at 991Mbit/sec. There's not going to be much more proof as to the inherent strengths of on-board NICs.

On-board NIC Challenge 2

Also, as crahak pointed out, a lot of on-board NICs have all the same features that expensive dedicated NICs do. I'm running a regular consumer motherboard, the ASUS P5Q Deluxe, nothing out of this world at $200 (certainly a lot cheaper then things like the Maximus or other high end gamer boards).

This on-board NIC supports

• IPv4 Transmit Checksum Offloading
• IPv4 Receive Checksum Offloading
• TCP Checksum Offloading
• UDP Checksum Offloading
• Large Send Offload
• 512 Transmit Buffers
• 512 Receive Buffers
• 30,000 IRQs per second
• 9014 Byte Jumbo Frames
• VLAN tagging and priority

and it's connected on-board through PCI-E ... what more could you want?

Wow, and i thought people here actually knew what they are doing... The Killer NIC is one overpriced marketing gimmick. It is cool if you want a mini Linux to play with inside your computer, but it won't magically lower your ping. And it would be $230 vs free, which is an unfair comparison in itself.



If you do not game any, and listen to music through $10 PC speakers, onboard will do fine. When you get a decent speaker system (and i don't mean Logitech) the difference is obvious.

If you looks at specs, they're very similar indeed.


Creative cards have EAX (for gamers), but that's about it. I find onboard codecs like Realtek's are far nicer and FAR more stable (Creative has a LONG history of driver issues). Modern onboard audio also tends to do a lot more little things too, like support for the codecs used by Blu-Ray discs, have 7.1+2 audio, have advanced features like DTS Connect no found even on Creative's most overpriced card, etc. Also, Creative cards tend to be braindead when it comes to interconnects (from proprietary plugs, to sharing the same jack for spdif and mic input and things like that). No need to waste $100 on a card either, and no filling PCI/PCI-e slots for basic functionality.


Not so. When you have a good speaker system (not one of those $100 5.1 setups), listen to high quality audio (preferably lossless), and also have a high-end card (not just any sound card -- LOTS of them are MUCH worse than most onboard audio), like perhaps a good M-Audio card, then sure there's some difference. On both my setups, my onboard audio beats my old $200 sound card hands down, that card is soon going to be donated.

The REAL place where people NEED sound cards (non-onboard) is mainly musicians, who need their fancy multitrack low-latency ASIO and such (and rich audiophiles, and gamers who can't live without EAX)

Here's one for on-board video. 87% of all video driver crashes last month were caused by nVidia drivers. That makes Intel GMA drivers pretty damned stable. And please don't ask for the source of this statistic, I'm not allowed to disclose that information.

That's what I was saying in the other thread too. They're rock stable. Never one crash, EVER. I wish I could say that about nvidia (or ATI even). And no 50MB downloads full of "extras" (like "helper services" and such things).

Even the Linux drivers for it are great (and it works just fine with Compiz and all). Whereas flglx (for ATI cards) on Linux crash X several times a day and has countless bugs.

There's nothing I need that it won't do. H.264 decoding acceleration would be nice but it's just that, and nowaday's CPUs have no problem decoding that anyway, and getting it to work with some video cards is nigh impossible (this geforce 8600 POS I have advertized the feature, but it never actually worked).

im usually pretty skeptical of statistics like that and use nvidia in any machine i can and i have no problem believing that statistic. Nvidias drivers have come a long ways but I can say that 9 out of 10 BSODs ive seen in the last few years had the text NV*.SYS on the screen.

That's where the importance of objectivity comes in. I myself use only nVidia video cards and adore them, yet I can still recognize the failure they are at writing drivers, unfortunately.

Don´t we love overlocking?

Most crashes I see are shaders/core/RAM out of sync (1MHz could mess up a card real bad).

onboard sound tends to suck though, you usually get feedback when you move the mouse etc

i do have HD595's though

I've only seen the mouse/video feedback issue on poorly shielded homemade audio amps.

I've seen that in a few rare occasions though I can't tell you what exactly caused it. I believe it actually has nothing to do with the mouse but it may actually have to do with something between video and sound. (Your video card does redraw the mouse pointer everytime it moves, doesn't it?)

I've had one problem with the Linux drivers. I was using a OpenGL based modeling tool (Comsol Multiphysics, to be precise), and with postprocessing of the display, the entire system would hang when running off the Intel ingrated video (this was a G965 chipset). The moment I tried an ATI or nVidia card in the PCIe slot - no hanging.

Still - that's the least problematic of the lot.

Well, we're talking about nvidia drivers here

They certainly don't need that to crash, and it's not just video drivers either. Most of the drivers they write seem to have issues (e.g. nvappfilter.dll -- even Mark Russinovich covered that one). This is why I don't buy boards with their chipsets (I've been burned by several -- VIA KT133 was the worst, but they gave up on making chipsets). Also, I remember seeing very similar statistics straight from MS themselves and surely 87% of BSODs don't happen because of OC'ing.


Ah, I don't think I've used anything OpenGL-based in lot of years, so I can't say.

Anyways. AMD's solutions are much nicer than Intel's here. The Intel chipsets with onboard video (e.g. G45) don't seem to make it to many boards (mostly cheap mATX), and AMD's is nicer in several ways (e.g. often has 1GB of sideport memory) and actually makes its way on some nice boards. Oh well.

Correct - especially if you're using Vista (and doubly-so for x64), watch out for nvidia drivers. I've actually seen the driver talking to an 8800GT completely hard-lock a Vista box when playing a video. I couldn't even crash it with a PS/2 keyboard or break in with a debugger on the COM port - completely, utterly hard-locked. Epic Fail.