AMD To Officially Adopt DDR500

[fbrdphreak]

Up until today, DDR400 has been the official memory standard for majority of the platforms, but Intel made the change to DDR2, and now it?s AMD?s turn to make a change to their platform. According to an insider source, AMD has plans to adopt DDR500 as its official memory standard in the very near future.

Before making the switch to other memory standards, the chipmaker has decided to extend support for DDR even further. The source said AMD would most likely support DDR500 with its Venice core microprocessors.

According to the information we received, AMD was in talks with memory makers to make this idea a reality since last year. In fact, the chipmaker was conversing about supporting low-latency DDR500 modules, as they tend to work best with AMD microprocessors and their onboard memory controller.

The insider said support for 250-memory divider is already available in numerous BIOS files, but it?s a matter of time before motherboard makers enable that feature. Since MSI supplies sample boards to AMD, all of their boards have the up-clock memory ratios that they can enable.

Thus far, that?s the only information we have received. All in all, AMD will continue to support DDR and may even move up to DDR500 as its official memory standard.

Link

:thumbsup:

 

[Sunbird]

I thought some JEDEC body decides whats "official" for the whole world?

 

[scsi drv1]

I'll finally be able to use my memory for its real purpose. YAY!

 

[Zebo]

Actually :thumbsdown: What that means it's hella hard, if not impossible, to OC a 3000 to 2500Mhz.

Let me clarify: they will release these chips with 250 HTT/FSB, meaning a "new" 3000 will have a 7 multi, stock Mhz = 1750.

Thier goal here I think is to get as close the mem controller max as possible and limit your OC ability.

To get to a measly 2500Mhz you now need (2500/7) 357 HTT!!!!

Not only impossible for most boards, the mem controller can't handle it

 

[RichUK]

Originally posted by: Zebo
Actually :thumbsdown: What that means it's hella hard, if not impossible, to OC a 3000 to 2500Mhz.

Let me clarify: they will release these chips with 250 HTT/FSB, meaning a "new" 3000 will have a 7 multi, stock Mhz = 1750.

Thier goal here I think is to get as close the mem controller max as possible and limit your OC ability.

To get to a measly 2500Mhz you now need (2500/7) 357 HTT!!!!

Not only impossible for most boards, the mem controller can't handle it

Damn ... good thinking Zebo .. i would not have thought of that, i would have been like "Oh that will be a great move for AMD, and even better for us OCers" ...

Q: with the mem controllers can they cause limitations on the latencies at those speeds ??

 

[Duvie]

I agree Zebo....BUT!!! If they do it just on a Bios level as they seem to be suggesting...(I dont think they want to rewrok another chip) then a person can decide to use it or not. It may be a very clear sign that the memory controller has been tweaked to run better at the higher speeds....

I think it may be a boom for non ocers who want to run a higher bandwidth at stock (limited performance gain actually) but for us OCERs we would tend to stick with whatever gave us the highest multiplier...

 

[Duvie]

Jus think of all of these CBBID stepping chips of late having a tough time corssing 240-245 HTT...Like I said this may be a sign of how much better the venice core (The only core they talked about in the thread) will be with its memory controller....

 

[Zebo]

They don't have to rework another chip D, They just cut traces for a 7 multi instead of 9 for 3000 , just a differnt cut. same cost.

This does' a couple things for AMD
a) about a 5% boost in benches at the tech sites w/o having to raise processor speeds due to increase Bandwidth assuming they keep LL like the article says AMD wants
b) kills the low budget OCers like US. You don't think they know we are out there OCing their $140 chips to FX-55+ levels

So:

7 x 250 = 1750Mhz they'll still call 3000
8 x 250 = 2000Mhz call 3200
9 x 250 = 2250Mhz call 3500 (actually more accurate this way)
10 x 250 =2500Mhz call 3800 (actually more accurate this way)

etc etc etc

 

[formulav8]

I highly doubt AMD cares about people ocing. I am sure it has nothing to do with that.



JAson

 

[bunnyfubbles]

Originally posted by: formulav8
I highly doubt AMD cares about people ocing. I am sure it has nothing to do with that.



JAson

Actually I think they do, allow all their chips to have access to lower multies and their FX series have access to the full range.

HOPEFULLY they don't force this on us and just have it as an option, it doesn't sound like something they'd do for their low end chips anyways.

Similar to the 400MHz FSB Barton 3200+ compared to all the other 333MHz FSB Bartons...it could be only for the high end chips. Zebo pointed out that the 2250MHz 3500+ and 2500MHz 3800+ would be more accurate (although they'd probably end up calling them 3600+ and 3900+ respectively), or perhaps a new non-FX flagship @ 2750MHz.

I find it hard to believe that they're going to move to 250 officially, especially across the board, and especially for single core.

 

[Maluno]

Originally posted by: bunnyfubbles

Originally posted by: formulav8
I highly doubt AMD cares about people ocing. I am sure it has nothing to do with that.



JAson

Actually I think they do, allow all their chips to have access to lower multies and their FX series have access to the full range.

HOPEFULLY they don't force this on us and just have it as an option, it doesn't sound like something they'd do for their low end chips anyways.

Similar to the 400MHz FSB Barton 3200+ compared to all the other 333MHz FSB Bartons...it could be only for the high end chips. Zebo pointed out that the 2250MHz 3500+ and 2500MHz 3800+ would be more accurate (although they'd probably end up calling them 3600+ and 3900+ respectively), or perhaps a new non-FX flagship @ 2750MHz.

I find it hard to believe that they're going to move to 250 officially, especially across the board, and especially for single core.

Correct me if im wrong , but I thought the reason for the locked multis was to prevent shady dealers from upping the multi and selling the chip as a different model.

 

[bunnyfubbles]

Well multis aren't locked on A64s, they allow the lower multis to be used. It is true that they are locked above their default, but you still have acess to the lower multis which are a great asset for true overclockers to get the most out of their system overclocks.

The FX series is an enthusiast series and with the prices they command it doesn't make sense for someone to sell them as a high end chips because they already are high end (ie it wouldn't be like taking a $150 3000+ and selling it for $500 as a 4000+)

 

[CheesePoofs]

The way I interpreted that article was that they would add a new memory divider (5:4) so that they can keep the HTT at 200mhz while raising the memory speed to 250mhz.

 

[Duvie]

Originally posted by: CheesePoofs
The way I interpreted that article was that they would add a new memory divider (5:4) so that they can keep the HTT at 200mhz while raising the memory speed to 250mhz.

That is the way I saw it as well....I think we should take this news as good...Menaing they have a better memory controller to run those certified speeds...And if AMD thinks 500ddr is certified then that means it is a lot more then that for us OCers to hit.....They said bios options.....I dont think AMD will try to confuse anything by introducing another chip with differing actual speeds. It would be hit or miss to get it and I think they would just want to offer all venice cores to get this support.....

A bios that either recognizes the Venice core and opens up a 250 ram divider or a blank bios option that at default is still 200 but you can change it to 250....My gut says the first one for OEM builders...

 

[DAPUNISHER]

Originally posted by: bunnyfubbles
Well multis aren't locked on A64s, they allow the lower multis to be used. It is true that they are locked above their default, but you still have acess to the lower multis which are a great asset for true overclockers to get the most out of their system overclocks.

That si just a nice bonus though, The lower multis are available because of C&Q and PowerNow! not for the overclockers benefit braddah

 

[Gogar]

I think the bios should just read it out of SPD and set a divider to 5:4

 

[ssvegeta1010]

Originally posted by: Gogar
I think the bios should just read it out of SPD and set a divider to 5:4

That what I think they mean. Itd be a good idea as async memory doesn't hurt A64 performance.

 

[jbh129]

Would running memory faster than FSB (I know thats the incorrect term) give any increase in performance? Seems unneccesary given the existing multiplier options.

 

[CheesePoofs]

It will give an performance boost, because yoru running the ram faster than you would be (250mhz compared to 200). The faster you run it, the bigger the speed boost you get.

You wont' get an extraordinarily large speed boost though, the A64's aren't very dependent on RAM bandwidth.

 

[Duvie]

I agree it should be setting HTT to 4x 250 and then leave ram at 200...otherwise asynch will be as wasted as pc2700 was on 266fsb Tbirds and Tbreds.....

However I say they should allow for the options but not have AMD set it so cause like Zebo said it will suck for us OCers with lock upper multipliers....