Computex 2006: 300W GPUs, Conroe, HDMI Video Cards and Lots of Motherboards
by Anand Lal Shimpi on June 5, 2006 10:24 PM EST- Posted in
- Trade Shows
The Memory Market
The memory manufacturers we’ve spoken to are ecstatic about AMD’s move to DDR2, as they can begin moving away from DDR.
While DDR shipments still outnumber DDR2 shipments, the shift is definitely taking place. Right now there is a price parity between DDR-400 devices and DDR2-667 devices; combined with AMD’s move to Socket-AM2 and the conditions are ripe for a serious transition over to DDR2 in the market. If Conroe ends up being as popular as everyone is expecting it to, the number of DDR2 platforms will continue to increase as will the demand for DDR2 memory.
The breakdown of DDR2 shipments is particularly interesting, with the majority of modules still being DDR2-533. Now that the price difference between DDR2-533 and DDR2-667 has gone down tremendously, we will see the balance shift more towards the latter.
Unfortunately DDR2-800 is still extremely expensive compared to DDR2-667, because of extremely poor yields. One source told us that DDR2-800 was yielding far below 50%, resulting in anywhere from a 30% to 100% premium over DDR2-667. Everyone expects the yields on DDR2-800 to increase tremendously later this year, which will help drive DDR2-800 prices down as well as increase the number of low latency or higher bandwidth DDR2-800 parts.
Samsung has been talking to its partners about the move to DDR3; by the end of this year Samsung is supposed to be sampling DDR3 modules, but the memory makers we spoke to believe that the schedule is too aggressive.
Currently on Samsung’s roadmaps we’ve got DDR3-800, 1066 and 1333, but things could very well change by the time DDR3 actually makes its introduction on the desktop. The memory makers we spoke to expect to see the beginning of a transition to DDR3 by the end of 2007 at the earliest, which seems plausible.
61 Comments
View All Comments
lopri - Tuesday, June 6, 2006 - link
Riiiiiiiiiiiiiiiiiight. I thought the next gen Intel chips are supposed to consume less power? I admit that I'm ignorant when it comes to such things, but the common sense tells me that VRM which supports TDP 130 should have no problem handling TDP 80? Could anyone elaborate?
DigitalFreak - Tuesday, June 6, 2006 - link
It happens every time Intel releases a new CPU. The then current chipset will support the CPU, but then the "VRM issue" pops up. It's a scam by Intel to force you to buy a new motherboard (preferrably with their chipset, of course). AMD doesn't seem to have this problem.ShapeGSX - Tuesday, June 6, 2006 - link
Yeah, slam that AM2 CPU into your old Socket 939 board. It will work! AMD doesn't require you to buy a new motherboard to buy their latest and greatest!Conroe is a huge departure from the P4. The fact that it works at all in LGA775 and with old chipsets is impressive.
phusg - Tuesday, June 6, 2006 - link
Bol(*#*&s, digitalfreak is right, of course this is a scam. Just because AMD is now using the same scam doesn't mean this is the way it has to be. I'm sure it wouldn't cost too many transistors to build in a legacy mode on the CPU die.Spoelie - Tuesday, June 6, 2006 - link
U guys are missing the point, the X2 dual core socket 939 cpu with TDP 89w works on the socket 939 motherboards that were available at launch 2 years ago, when .13µ newcastle cpus were the only game in town, barring any bios updates of course.Same thing for socket A, altho not at the right FSB speed depending on chipset.
AMD only forces you to update if there is a significant feature difference that can't be worked out in the current socket. It seems that intel engineers its cpus without 'socket environment' consideration and then engineers a specific chipset for it, while amd engineers its cpus to fit in a specific socket environment that they defined a standard for long ago.
ShapeGSX - Tuesday, June 6, 2006 - link
So you don't think that lower power consumption due to the new VRM is a significant feature difference? It is all the technical press is talking about these days.The LGA775 socket was launched in June of 2004. So, here we are 2 years later with the same socket for Intel, but a change to the voltage regulator. And you can run dual core processors on what was originally a socket meant just for one core, just like AMD.
Socket 939 was also launched 2 years ago, and it is already obsolete.
Sounds like both companies have similar track records, as of late. Perhaps the march of technology these days simply will not allow for using the same motherboard for more than two years.
JarredWalton - Tuesday, June 6, 2006 - link
It could be that the VRM requirements are for lower voltages or cleaner power or something along those lines. Also, just because Conroe runs cooler than Presler doesn't mean it can't have more stringent voltage requirements. I wouldn't be surprised if this is less of a case of *can't* run Conroe but more can't run 100% *stably*. The newer 975X chipsets/motherboards will probably have a few slight tweaks to fix some erratta encountered with current 975X designs.giantpandaman2 - Tuesday, June 6, 2006 - link
Conroe has new power states that aren't supported by older motherboards. For example, a Conroe would try to go to "sleep" and the motherboard wouldn't supply the right level of power to it.soydios - Monday, June 5, 2006 - link
1 kilowatt power supply? bloody hellI like all the pictures. But I'm still waiting for Asus AM2 RD580.
highlandsun - Monday, June 5, 2006 - link
Any news on iRAM2 or anything similar?