Abit it7 motherboard downloads

Lower latencies are of course A Good Thing. So its extra bandwidth. Why that is, is because of the latency and the cited article takes you through and shows you exactly how real world bandwidth is derived. It was purposeful to show you that negligible difference as it shows a distinct point.

Then incorporate what you have learned about latency and it impact in creating real world bandwidth. The answer fits with your quoted segment of the article. In fact we know now that high latency lowers real world bandwidth and that peak bandwidth is just a pretty number. The reason I included the date the article was written was to empthasize the fact that it was done with the PIII in mind, not the P4, which is the processor the arguement centers on.

The result corroborates that latency x clock results in the real world bandwidth of the memory platform. This consequently why Brookdale was so poor a choice for the P4.

Not because it didn't have great latency, but because it simply could not provide enough real world bandwidth to keep the P4 busy. In other words, the P4 sits around twiddling it's thumbs. And as displayed in the articles chart: PC Latency: You have to lower the latency and increase the clock and then you have a positive growth. PC Latency: This is a old and constant battle to keep latency from rising and making the clock speed higher.

Because it's real world latency penalty x clockspeed is not enough to generate real world bandwidth in gross excess of what the P4 can use. Since the delta in performance is so small, the outcome of the arguement does not translate into very useful information on how to enhance performance by any significant amount.

I tried to lead people into the conclusion versus just laying it all out. I should have just made the point clear to begin with instead of just linking. It's right on the money. It's called Random Access Memory for a reason and the longer the penalties to get the goods out of it the worse your real world bandwidth is going to be.

Your reponse is:[quote]The results from that formula create clear bandwidth results which have been coroborated by modern stream results, even on the P4. It addresses latency and the impact that those strobes have upon bandwidth. In fact as the article details real world bandwidth is not determined by width by clock, but instead with latency cycle by bus cycle time. The results from that formula create clear bandwidth results which have been coroborated by modern stream results, even on the P4.

Unfortunately I feel that the material I provided did a wonderful job of clearly detailing the exact point I was trying to make.

You can substitute the word software for cases if you please but that was the meaning of the statement. The P4 is not a bandwidth only machine. My amended statement still holds water. Unless you use CL3, go for a higher memory speed rather than lower latency.

Take CL2. It works! How, when all the chipsets being tested are Intel chipsets? I was, too, until I switched to generic white heatsink compound instead of epoxy, and window defogger repair kit instead of silverized epoxy crap.

The above situation does not remain constant and certain software looks for one thing over the other. Nonetheless it does illustrate the the P4 is not ruled by bandwidth, and it will take latency over bandwidth when the situation calls for it. Here's another example. I never stated or cited any other OS. If I was unclear on this point I apologize. So are we creating a bit of a chicken and egg here. More bandwidth helps, but obviously lowering latency does too. In the first link they are using the exact same memory speed for each latency setting.

It is a no-brainer that lower latencies at the same speed improves performance. Your second link ties into what I was trying to say. The chicken-egg thing is a good analogy. Both do good things for memory bandwidth but on the P4 latency is less important than burst memory bandwidth.

That is until we saturate the Mhz FSB. The Pentium 4 gets faster when lower latency is applied. Latency and bandwidth are equally important and one does not have benefit over the other even on the P4. Rambus works in this exact fashion which is why we see such impressive numbers despite such high latency. Neither of those articles were written in the context of the Pentium 4 architecture which is my entire point.

The P4 benefits from sending larger chunks of data down the bus despite higher latency. Windows XP was used in this review. Also, most of the issues listed in that PDF were fixed by patches from the manufacturers of the programs not Intel. You cite a link to an article using Windows ME.

Windows XP is the OS in question here. Which is odd considering what items are benchmarked within that test. Even under Windows XP there are performance benefits and bug fixes to be reaped from installing the 4in1 drivers.

At the same clock, yes. However, in this case sacrificing raw Mhz for lower latency is a bad decision. The Pentium 4 will perform better with higher memory speed and higher latencies rather with lower memory speed and lower latencies.

Being able to send more data per clock despite longer times between clock cycles is more important. I was trying to be a tad sarcastic. Why would you test an Intel chipset without their drivers? The IAA is not necessary and if put under heavy scrutiny may prove to be nothing more than a clever cheat like their old Busmaster IDE drivers.

Whether this is the result of poor support from Microsoft, initially poor efforts by VIA at writing drivers, poor efforts at VIA making hardware, or an intensely clever plan to leverage constant driver updates in order to influence benchmark scores positively in new applications I'll leave to the conspiracy theorist to figure out.

There is either something wrong with the motherboard or the person who set it up. I do think it hurts the credibility of the reviewer when such, uh, questionable results are taken at face value. Diskbench 99 says so! Highpoint rules!! I seem to recall a lot of vendors optimizing for specific benchmarks, and this one is fairly old.

Because quite frankly, I think those disk results are ludicrous. There is no way the highpoint controller is 2x faster than the Intel one— the bottleneck is the hard drive and the interface, by far. In fact, its whole design is built around high latency memory Rambus.

It wants raw bandwidth. Screw fast timings and crank the bus speed. All I would add is that, since about 2 years, Abit mobos left me the undesirable feeling not to be reliable on the long term 5 or 6 corpses so far. Regarding Matrox Parhelaugh, I strongly agree with you. Is everybody expecting so much of upcoming drivers?

Same thing for the oh-so-wonderful P10, with this revolutionary design which only led to revolutionary benchmarks : the brand new card is much slower than half-priced cards…. Last but not least, integrated peripherals benchmarks would have been nice.

What performance is lost from using the onboard sound? Same thing about the Ethernet controler. What about some less questionned HD bench than the infamous Winbench? Whats goin on here?

One of the reasons I buy Intel is to avoid the hassle of Via 4-in I just want the it to work out of the box. I think a lot of people out there just use whatever drivers that comes with Windows. What in the hell is going on? Next gen? Were they even tested? How about the performance of the integrated audio and NIC? I have been telling people about this forever.

They were supposed to make any hard drive and controller super duper fast in Windows. And they had benchmarks and older version of Winbench to prove it. However, when the then-new version of Winbench came out, performance with the goofy-ass Highpoint drivers totally sucked! But they never made similar drivers again. Now, Highpoint has since gone on to make actually useful hardware. Was the built in controller tested with our without those installed?

Unreliable and slow and very limited space. As you can see, we have a lot to cover. Given the limited availability of even high-latency DDR memory, low-latency DDR may offer better overall performance, anyway. Scrapping legacy serial and parallel ports gives Abit plenty of room on the port back plane for loads of audio, USB 2.

You can access all but one Firewire and two USB 2. At the very least, it probably simplifies tech support for them a little. The Highpoint chip controls drives plugged into the Serillel adapter. Unfortunately, the connection between the Serillel adapter and every hard drive I tried it with was so loose, even a gentle bump could cause the adapter to fall out.

This board should appeal to enthusiasts more likely to move their cases to and from LAN parties, so a loose hard drive connector is particularly annoying. Fortunately, Abit gets it right. I may just be a sucker for zip ties. The tube of odd zip ties on my work bench would support that theory. The cable binders made short work of the mess of CAT5 cable running back and forth between my hub, test machines, and server. Front-side bus speeds are available up to MHz in 1MHz increments, and there are plenty of multiplier options should you be lucky enough to have a Pentium 4 chip with an unlocked multiplier.

A menu choice also allows users to lock down the PCI bus arbitrarily at 33, 37, or 44MHz, regardless of the front-side bus speed. Memory voltage options are a little more limited than I expected, but you can still take your DIMMs from 2. The CPU voltage options are considerably more abundant—sort of. CPU voltages go down as low as 1. Oddly, the CPU voltages top out 1. You might expect to see more tweaking options than the ones in the screenshot above.

Generally, however, there are fewer tweaking options available on Pentium 4 motherboards, especially those based on Intel chipsets. Not only can you set automatic shutdown and warning temperatures for the CPU, you can also have the system sound a warning or power down if your heat sink fan fails.

The fan and thermal shutdown features should protect against most such disasters. The primary difference between the two, performance-wise, should be attributable to the additional memory bandwidth provided by the MAX2 V2. Vertical refresh sync vsync was disabled for all tests. All the tests and methods we employed are publicly available and reproducible.

If you have questions about our methods, hit our forums to talk with us about them. In Business Winstone, the extra memory bandwidth is worth less than a point—nothing earth shattering. So which option offers the best performance? Curiously, using the Serillel adapter gives performance a solid boost, despite the fact that the drive is still interfacing with the HPT controller.

Any data that CRC check discovers is bad must be resent; resend data and waste time. SATA signal is so clean, there is basically no need for data resends. This effect is so great that in effect it easily makes up for the double conversion. Good question. R2P2, my point exactly. Abit is just using 2 Marvel chips to transform the parralel protocol into a serial one and back.

Something is fishy somewhere…. This is getting confusing. AG19, well put. It seems the public are in love with labels though, regardless of what these labels bring them. No major slowdowns in Win2k… the system appears to be just about as fast as before when I was using gb JBs connected directly to the onboard.

Abit contributes in one way, but eliminating the legacy ports, but Apple and Intel still duke it out over their competing standards. If only it were as easy as slapping an adapter between the various things that require serial and parallel ports. Progress that is continually leaving people behind. I hope we can expect to see lots of Serial ATA peripherals out in force soon. But something tells me it will be before it picks up in a big way.

Another good review, even though the board doesnt do much for me. Type search above and then hit Enter. Now, Abit is focusing on integrating loads of peripherals instead of pushing a legacy-free agenda. PCB real estate for standard ATX motherboards is limited, which makes squeezing all those chips and ports onto a board especially challenging. A less offensive shade of brown. Only three DIMM slots? Serial ATA today, almost.