Athlon 64 3400+ Motherboard...

[Akashio]

Does anyone know of any Socket 754-pin Motherboards capable of carrying an AMD Athlon 64 3400+ that has potential for future upgrades.

I found Asus K8V SE Deluxe--tell me what you think:
http://www.computerhq.com/ASUS_K8V-SE_D ... 47983.html

How do I know if a motherboard has dual-channel DDR capabilities? BTW, isn't twin ddr (2 x 1024MB) faster than a 2048MB stick by itself?

[Scintilla]

Does anyone know of any Socket 754-pin Motherboards capable of carrying an AMD Athlon 64 3400+ that has potential for future upgrades.

Last I checked, none of the 754s had much potential for future upgrades, because AMD's pushing Socket 939 -- all the CPUs they have planned for 754 for the future are Semprons.

Unless you're talking about other upgrades. But it's still something to consider.

[Akashio]

so freakin expensive tho... Do you think it's worth it?

[Scintilla]

I do not feel qualified to give a judgment on that, as I've never used an A64 machine. You mentioned upgradeability, so *shrug*
Try asking DWChang or Trythil...

[Brolly345]

so freakin expensive tho... Do you think it's worth it?

I'd say no because if I remember correctly AMDs are not so good at handling floating point equations. They're only good for arithmetic equations, which are not really what a lot of advanced programs use.

Make your own decision though.

[dwchang]

so freakin expensive tho... Do you think it's worth it?

I'd say no because if I remember correctly AMDs are not so good at handling floating point equations. They're only good for arithmetic equations, which are not really what a lot of advanced programs use.

What?

Where are you getting this?

If you're like a Computer Engineer or the engineer who designed the FPU for Intel then I'll believe you.

FPU's are difficult to design, but there is no major difference between the two processors. That you can see from YOUR HUMAN point of view. They go about solving things different, but get the same result.

If anything, I'd say you're actually wrong in the opposite direction. The floating point pipeline for a P4 is something close to 40 stages while the Athlon series is well below 20. At the same time, the P4 runs at a faster frequency per stage (i.e. it does the stage faster) so it evens out TO SOME DEGREE.

If anything if an Athlon-64 had a 20 stage deep FPU pipeline and the P4 a 40, that means the P4 would have to operate at twice the frequency to derive similar performance.

Anyway back to the original post, I strongly suggest a 939 board over a 754. Take the $$$ hit so you can upgrade in the future. We (AMD) has stated that the 754 will be gone by 2006 (although the Sempron may stick with 754 pinset) and the Athlon-64 set will be 939 only (with 940 as server). Just spend the extra dollars, go 939 and in a few years, go dualcore 939

[trythil]

so freakin expensive tho... Do you think it's worth it?

I'd say no because if I remember correctly AMDs are not so good at handling floating point equations. They're only good for arithmetic equations, which are not really what a lot of advanced programs use.

Yeah, because floating point isn't arithmetic.

You probably meant "integer performance", although I still want to see proof for what you state. Also, floating-point performance is just one aspect of overall system performance for video editing. It's not something that you'd base a buying decision on by itself.

[sysKin]

I'm quite sure no part of video editing is floating point. Definitely decoding (both mpeg-based and huffyuv) have no floating point at all, filters don't have it at all, and encoding doesn't have it at all. And GUIs either.

Radek

[Brolly345]

Well, I did some research on this subject. Because I was considering going AMD, and I found out some useful information from a thread here on AMV. Maybe some of you remember it. It had a picture that showed the benchmakrs for the latest Intel and AMD processor. Thread reference here. Note the picture is no longer there, but it was enough to make me go looking.

I found a lot of information just using google on the definitions of Whetstone and Dhrystone. I found a lot of web pages, forums, ect on the subject. All of the sites I found that were related to the benchmarks told me that Whetstone is floating point, and Dhrystone is arithmetic, or string, or whatever you want to call it. So now if any of you can remember that picture from that thread that showed the results of the benchmarks on the AMD and the Intel, I should have saved that picture, know that the AMD only slightly outperformed the Intel on Dhrystone, but got killed by the P4 in Whetstone.

[trythil]

I'm quite sure no part of video editing is floating point. Definitely decoding (both mpeg-based and huffyuv) have no floating point at all, filters don't have it at all, and encoding doesn't have it at all. And GUIs either.

Radek

Some video editing and compositing systems can make use of floating-point RGB data when performing composition or effect work; Cinelerra is one such system.

That's about all that's coming to mind, though...

I found a lot of information just using google on the definitions of Whetstone and Dhrystone. I found a lot of web pages, forums, ect on the subject. All of the sites I found that were related to the benchmarks told me that Whetstone is floating point, and Dhrystone is arithmetic, or string, or whatever you want to call it. So now if any of you can remember that picture from that thread that showed the results of the benchmarks on the AMD and the Intel, I should have saved that picture, know that the AMD only slightly outperformed the Intel on Dhrystone, but got killed by the P4 in Whetstone.

If you read that much, you should also know:

- Whetstone is a synthetic benchmark. (That alone should raise some suspicion if you know anything about how these things work: in particular, that this is generic code not specifically scheduled for a particular processor's architecture. Compilers can do more and more with each month, but if you really want to squeeze the best performance for an algorithm, you're probably going to have to do it yourself.)

- Related to the first point: There are different versions of the Whetstone tests! For example, there's the "generic" version, and then there's the version optimized using the SSE2 instruction set. It's obvious that proper usage of SIMD stuff like SSE/SSE2 will allow a significant speedup (in best case, proportional to the number of data units that can be operated on simultaneously) on processors that implement it. Which graphs were you looking at?

- Whetstone tests both floating-point and (to a lesser extent) integer performance.

- Arithmetic operations are not string operations. (Actually, most processors don't have instructions to operate on strings.)

- If you really like pictures, this one demonstrates the dual Xeon getting spanked in most of the Whetstone tests. Which Whetstone? Who knows, they don't say. Might as well be worthless results.

From this, I'm not sure how you cannot draw the conclusion that results from any synthetic benchmark, without context, are total garbage.