Sign in with
Sign up | Sign in
Your question

today's memory article

Tags:
Last response: in Memory
Share
July 1, 2003 3:47:33 PM

Quote:

But DDR RAM doesn't really spread its wings until it's installed in dual-channel motherboards that add together the memory bandwidths from two DDR modules. These mainboards sport Nvidia nForce 2 chipsets for AMD CPUs, or Intel 7205 (and, in the future, 865 and 875) chipsets for Intel processors.

I guess this isn't a recent article. I'm not sure the references to SPD settings are accurate for 865 boards trying to activate PAT either.

More about : today memory article

July 1, 2003 4:38:44 PM

From what i have seen on amd motherboards...dual channel ddr on the nforce2 doesnt provide that much performance...not enough for me to use the words "Spread their wings."
July 1, 2003 7:40:24 PM

Quote:
Long gone are the days of Windows 98 and Me, when 512 MB was the most memory that the majority of systems needed.

That is a moot statement. Windows 9x (Chicago) could not support more than 512MB, anyway, without serious problems or third party hacks.

__________________________________________________
<b><font color=red>Three great virtues of a programmer are: laziness, impatience, and hubris.</font color=red><b>
Related resources
July 1, 2003 7:49:56 PM

This article is less extensive than what I expect from Toms!
-For one, there's no conclusion in the end!
July 1, 2003 9:03:06 PM

I 2nd that. must to busy with the media black out lol didn't answer anything i was looking for just confused it more. looks almost like they read a motherboard manual and copied it to the web.
July 1, 2003 9:23:45 PM

I would like to get the amazing timing that they had though

*drool*
July 2, 2003 3:27:20 PM

I've noticed a couple 'typos' in there, they are as follows:

- <A HREF="http://www6.tomshardware.com/howto/20030701/memory_tuni..." target="_new">Content Creation Winstone 2002 Benchmark</A> shows PC2700-1.5-2-2-5 with two different scores. Which is the right one?

- <A HREF="http://www6.tomshardware.com/howto/20030701/memory_tuni..." target="_new">VIA KT400A</A> is listed as having a Dual-channel memory interface. AFAIK VIA doesn't have a chipset that supports Dual-Channel memory.

I sent an e-mail to comment@tomshardware.com about them, hopefully they will be looked into.

<A HREF="http://www.anandtech.com/mysystemrig.html?id=24106" target="_new">My System Rig</A>
<A HREF="http://service.futuremark.com/compare?2k3=535386" target="_new">3DMark03</A>
July 2, 2003 7:51:53 PM

I'd like to see 1.5-2-2-5 at PC3200
July 3, 2003 6:07:30 AM

FallOutBoyTonto wrote:
"I sent an e-mail to comment@tomshardware.com about them"

I tried to send a comment to comment@tomshardware.com, but the message was rejected - "The email you sent was considered unsolicited SPAM and was not delivered!
We're not interested and we don't want to get any future mailings."
July 3, 2003 1:42:58 PM

I have to say, I was <i>really</i> unimpressed with that article. It contained an awfully large amount of misinformation.

Quote:
older DDR-1 technology runs up against 400-MHz memory clock speeds like a bull against a brick wall.

You more or less have to assume that based on the context of the paragraph that this quote came from that it was in reference to DDR400 AKA PC3200. A bull against a brick wall though? Gee, we only have it running on numerous chipsets from several different companies with minimal problems. The only memory controllers that come close to having problems with DDR400 are dual-channel controllers, and even nVidia's seems fine now. So we're really just talking about Intel then. And even then DDR400 works just peachy-keen with normal DDR400. It's only the special low-latency sticks that are having problems actually running at their rated abnormally low latencies. So a bull hitting a brick wall? Only if the bricks are made of styrofoam.

Quote:
You can start out with special tweaking modules that exceed the DDR400 standard and offer particularly high settings for clock speeds and timings. The modules are available from Corsair, Geil, Kingston, Mushkin and others as PC3500 or PC3700. While standards by these names don't really exist, the names indicate how much the modules can be overclocked.

Sorry, but the module <i>isn't</i> being overclocked if it is being run at it's rated speed. If you buy PC3700 and run it as PC3700 then that's stock, not OCed. They might be useful for people who want to OC their CPU, but they aren't OCed themselves if they're run at stock.

Quote:
These mainboards sport Nvidia nForce 2 chipsets for AMD CPUs, or Intel 7205 (and, in the future, 865 and 875) chipsets for Intel processors.

In the future? One has to assume that:
1) This was just an old article that was only recently released. (Likely an old article that was only recently translated to English at that.)
2) That the editors involved (if any) either have no technical knowledge or never actually read the article before posting it on July 1st.

Quote:
BIOS often puts the brakes on the RAM timings so that the system remains stable, and that's where most of the optimization comes in.

To my knowledge VIA is the only one guilty of putting brakes on timings, and then only for DDR400 with early chipsets. Please correct me if I'm wrong, but unless you 'tweak' (a nice way of saying <i>overclock</i>) your RAM, BIOS will just run the RAM <i>at stock</i> according to it's SPD. Since when is 'stock' considered putting "the brakes on"?

Quote:
High-quality DDR333 RAM with quick timings will outperform a DDR400 module with timing settings that have been deliberately rolled back to increase the clock speed.

For an AMD system with a 166MHz FSB or lower, mostly (but not always) yes. For an AMD system with a 200MHz FSB, I have no idea, but probably not. For an Intel system, generally <b>hell no</b>. The P4 cares more aboud bandwidth than about timings.

Quote:
When tweaking your memory, the first step is to deactivate the automatic RAM configuration. When this function is activated, the mainboard reads the SPD chip (Serial Presence Detect) on the memory module to obtain information about the timings and clock speed and to adjust the settings accordingly. However, these settings, which the RAM manufacturer stores in the EEPROM chip, are very conservative in order to ensure stable operation on as many systems as possible.

Why is the word <b>overclocking</b> not mentioned once here? We 'tweak' our memory. How quaint. Because it isn't like that doesn't invalidate a waranty just as quickly as OCing a CPU does.

THG has been notorious (in a good way) for these kinds of warnings in the past, yet here I feel like we're preparing tea and crumpets. Let us now 'tweak' our RAM beyond the manufacturer's specifications. Would you like some cream with that?

Quote:
To see for yourself how big this impact is, take a look at the performance benchmarks for MPEG-4 encoding.

Benchmarks that tell us something useful are great. Is this single-channel or dual-channel? Is this on an AXP or a P4? Which chipset is this? Besides the typos in the charts and the strange lack of any speed of RAM other than PC2700 being used, there is the <i>very</i> peculiar lack of any information whatsoever on the hardware and firmware involved in this benchmarking. Are we just meant to believe that all memory timings on all chipsets for all CPUs adhere to these findings? How are these benchmarks even useful when no information is provided to help us use them?

Quote:
VIA KT400A | DDR400 3)

Wow! A VIA KT400A is a dual-channel memory controller now! I can't wait to tell VIA this!

Quote:
Intel 865P | DDR333

Crap! And now the dual-channel memory interface for the 865P has just up and vanished! I better go send Intel a letter of complaint...

Sadly, these aren't even <i>all</i> of the mistakes and lackings in that article. These are just the most blatant. There is no better label for this article than just plain crap.

I don't mind an OCing article (actually I love them) when it's made obvious that you're OCing your hardware and that this invalidates your waranty. This however was not such an article. The theme of this article was to shame people who run their PC hardware at stock (or if you're being generous, to intice these people) to invalidate their waranties by overclocking their hardware <i>without</i> even so much as warning them first. That is a <b><font color=red>BIG</font color=red></b> red-letter no-no for which THG should be ashamed of themselves.

Further the article is just chock-full of misinformation in general and a lack of useful (expected) information for the benchmarking that was done. I haven't seen such a lack of professionalism from an article at THG in <i>years</i>. Benchmarks are <b>always</b> preceded by system information. Benchmarks are also <b>always</b> a hell of a lot more thorough than that whenever conclusions are being drawn from the benchmark results. And on top of that all, again, other forms of misinformation abounded throughout this article. So again, THG should be ashamed of themselves for even so much as posting this article.

It's just sad. People make mistakes, but this is one mistake that can most definately be corrected. This article should <i>never</i> have made it to print at THG and it certainly should be retracted and/or corrected now that it has.

"<i>Yeah, if you treat them like equals, it'll only encourage them to think they <b>ARE</b> your equals.</i>" - Thief from <A HREF="http://www.nuklearpower.com/daily.php?date=030603" target="_new">8-Bit Theater</A>
July 4, 2003 12:47:13 AM

Yes, that quote was from the returned message.
I sent the returned message to postmaster,
he replied that my message just happened to be wrongly trapped by a spam filter, and that the filter has been adjusted.
I recent my email, this time it was not rejected.
!