MSI Lies about their Multipliers on the K8N Neo 2 Platinum?

[Npherno]

Hey all. Found this on HardOCP. Interesting article. It would help explain a lot about the design of our "enthusiast" boards. I sm kind of sorry I didn't buy a VIA chipset design now.. Here is the Link: http://www.sudhian.com/showdocs.cfm?aid=616

I have a similar situation where I have a 3200+ Processor, 90nm, Scythe FCS-50 cooler, K8N Neo2 Platinum BIOS 1.30 and a Maxtor MaxLine 300 Gig SATA disk. I have a OCZ 520 Powerstream, and 3200 Platnum Rev.2 memory.

When I first set up the board, with defaults, I couldn't get Windows to install. After babnging my head against the wall, I figured the memory wasn't timed properly, so I set it to the Factory spec of 2.75 1T-2-2-2-5 and the machine still wouldnt boot. It was only after I found these forums that I concluded the board couldn't even come close to handling those specs, and I am timed at 2T-3-3-3-8, to be stable at all. I was only able to push the CPU to 220 FSB, 5HT.

Now it was my impression that so long as you had the correct memory, good power, and adequate cooling you could get a good OC. Guess again! Anything over 220FSB, 5HT, the board defaults to "safe mode" and loads the last successful settings.

Does anyone have an idea of a GOOD 939 board with features that REALLY work?

PS I have done ALL the tests on these forums (moving memory, clearing CMOS, moving SATA connections, memory testing, etc.) My CPU and components are solid.

-N

[cool 'n quiet]

Let me just say that there's a lot of uncertainty about the sudhian review, and the jury is still out. Most people who have flamed in one way or another have missed the point: they note that memory speed actually runs below DDR400 when using a half multiplier, due to whole number dividers. True enough, but that's not what Sudhian was claiming. The claim is that the HTT is being artificially boosted, which should be something that can be verified just by looking at a cpu-z screen (yet Sudhian didn't do that).

I think that the MSI board works, but you need to be aware that it's made by MSI. That means it's going to be the board you love to hate, and hate to love. Translation: quirky, but if you avoid the wrong settings, and use the right ones (and stand on one foot and only use the board during a full moon) then the board will kick butt. Otherwise, wait for DFI to release a 939, or for ABIT to get out the funk they've been in the past few years since Oskar Wu left.

More seriously, there are a growing list of known NF3-250 chipset bugs that cannot be fixed by a mere bios update. This may be one of them.
Others include:
HT 1x and 2x won't POST. 2.5x, 3x, 4x, and 5x are fortunately OK.
SATA 1+2 have no lock. Do not under any circumstances use if your HTT goes near 230.
AGP Fast Writes hurts nvidia graphics card overclocks by at least 30mhz (debateable whether this is a bug, but many think it is since it doesn't seem to occur with other combos of nf3 vs. via, nvidia vs. ati)
Plus a half dozen more identified by Oskar Wu (posted somewhere on xtremesystems but I forgot where. If you really want a complete list, you can search there...)

Price competition has caused motherboard quality to become crap. Only one company, Gigabyte, is bucking the trend and selling motherboards for "high" prices (which would be worth it, if they work). But the jury is still out on Gigabyte's motherboards. The high prices may just be for features like their double speed firewire, not overclocking ability.

[SteveOCZ]

Interesting :hide:

MSI we know your out there. Shed us some light on this matter :)

Now while this does pertain to OCZ memory as well I'm going to move it to the AMD motherboard forum.

Hope you don't mind.

Steve

[Peanya]

Wow cool 'n quiet!!! True words so many should pay attention to! Remember the MSI 655MAX? That was a big name for the P4 before the 875 chipset came out. The Neo2 is extremely popular, but that is because it was the first available for the 939, and the glowing review it got at anandtech.com. I considered this board myself, but my worries were the nightmares I heard of, even before this article came out.
Who's to blame? Nvidia? The Nforce2 Ultra was a stellar chipset, but ever since then, their newer products were just lackluster. At least they have the 6800 series video cards to keep them looking good. MSI? I've heard many stories about their quality varying greatly.
Abit has been lacking in the OC department for a little while IMO, but they still have the benefits of a modern production facility, and very good quality parts. Maybe Asus will step back in the game too. Heck, I remember when a cheap motherboard was like $200... now people whine about a board costing $120 :scratch:

[mad mikee]

1. 220x5 = 1100MHZ > 1G for HT, drop to 4 and prolly do better

2. SATA 3,4 (near AGP) Okay for HTT > 240, NOT 1,2 (Thanks Oskar).

3. 1/2 multis, just say NO.

4. Changing Multi = reboot, as soon as you see post screen, POWER OFF (at PSU), then restart (Also Thanks to Oskar)

Hopefully these should help a bit.

[BB_One]

Maybe Asus will step back in the game too. Heck, I remember when a cheap motherboard was like $200... now people whine about a board costing $120 :scratch:


Hang on to your hat there, for the new ASUS coming out soon, anticipated retail price:

$347.00 ~ $376.00 REAL dollars not the Canadian monopoly money ! :yikes:

Now if that is not crazy what is? :coocoo:

http://www.xbitlabs.com/news/mainboards/display/20041029020001.html

Edit: Well I do admit that $800.00 + processor is crazy nah, that's insane ! :bonk:

[prometheus]

I first published info on the half-multiplier problems in my Athlon 64 Memory review over a month ago. I discovered it when I was testing the DFI.

Below is my letter to Joel at Sudhian sent a couple of days ago:

"Joel -

This is not new news. If you will check p. 13 on my first AMD memory test article published on September 21st at http://www.anandtech.com/memory/showdoc.aspx?i=2215&p=13 you will find a table with corrected memory half-multipliers and the following comments in Italics:

During testing, we discovered that while half-ratios do reflect CPU settings accurately, they are not really linear for memory speed. Memory speeds at full ratios are reported accurately. The asterisk values are the actual memory speeds at half ratios - not the reported memory speeds. Special thanks to Oskar Wu of DFI for his help in determining true half ratio memory speeds.

Oskar had been working on this info for a couple of weeks prior to publishing the article. I also disagree this is any kind of intentional deception on the part of MSI, since DFI encountered the same issue in designing half-multipliers on their 754 Enthusiast board. If you look at how the half multipliers are implemented it is easy enough to understand. The CPU half multipliers are correct, but the memory definitely is not.

Wesley Fink
Senior Editor, AnandTech, Inc.
wesley.fink@anandtech.com
www.anandtech.com"

Joel has graciously responded back and indicates he intends to do more inverstigation. This is not a conspiracy with MSI, it is an issue that anyone trying to implement half-multipliers on A64 is struggling with.

[RotorHead]

Hey prometheus, long time no see.....

Thanks for sharing that info and shedding more light on the subject. :thumb:

hehe....don't let em work ya to hard over there :beer:

[BB_One]

Seems it did not take long for Sudhian to respond prometheus.....

http://www.sudhian.com/showdocs.cfm?aid=617

Edit: Very interesting arguments, and some back up material to support his claims. This one is definetly going to be an intersting one to follow:

Excerpt:
The fundamental mis-assumption made by Anandtech, enthusiasts, and other sites was that the half-multiplier setting of the Athlon 64 actually existed or had been manipulated in any way...........As we expected, increasing the HTT clock yields a proportional increase in the memory speed. Everest predicts precisely what our table does. CPU-Z is wrong.

[BB_One]

Oh boy, man oh man do I not want to change my shoes with Joel's shoes at Sudhian at the moment.....

Seems one off the main element he was using for "swearing to the GOD's" that there is no multiplier on AMD 64 cpus, and also saying the CPU-Z is wrong", is the Everest Benchmark Tool.

Poor Joel, got a post on his forum that say's something like this:



Hi Joel,


This is Tamas Miklos, CTO of Lavalys, Inc. We develop Lavalys EVEREST that you used during your investigation.

I'm afraid EVEREST was actually wrong about the multipliers and memory clocks, and CPU-Z was right about all the multipliers and clock speed.

You can read the full thread here (http://forums.sudhian.com/messageview.cfm?catid=73&threadid=67129) , if you follow on various forums, seems from the start Sudhian was getting beat upon, and now..well.....overall credibility might be at stake here.

[BB_One]

Well, a small update on this matter which I guess closes it all together. One has to give credit and recognition to Joel for being man enough to 'publicly' recognize his mistakes and provide apoligies. However, he does bring to light a few good points I believe.

Update (http://www.sudhian.com/showdocs.cfm?aid=633)