AMD Motherboards: Processor Support
by Wesley Fink on June 28, 2005 1:00 PM EST- Posted in
- Motherboards
Final Words
The latest AMD Athlon 64 processors bring with them performance improvements, and a new round of potential compatibility issues. Venice, San Diego, and Toledo (x2) all have the 90nm manufacturing advantage and the addition of SSE3 instructions. While they will all potentially work on any Socket 939 board, all three new processors require a BIOS update, which may or may not be available for your favorite board.In general, motherboards based on the nForce4, VIA K8T890, Uli1689, SiS 756, and ATI Crossfire AMD will likely already have a BIOS released that will work with x2 processors or one will be coming shortly. Those with motherboards based on the earlier nForce3 chipset may not be so lucky - as BIOS' that support x2 on nForce3 are just starting to appear. We are told nF3 support for x2 is coming, but it is definitely coming more slowly. This means most Socket 939 boards with AGP video may not be supported right now with an x2 BIOS upgrade. We say most, because the new ULi 1689 chipset does support both AGP and x2, and some of the older but popular VIA AGP boards have x2 BIOS updates available.
AMD has also added additional "hidden" features in the AMD on-processor memory controller. Additional asynchronous ratios are available at 433, 466, and 500 memory speed on boards that implement the necessary code to access these memory controller features. These options should be available with any Revision E chip if the manufacturer implements the controller option read in BIOS. We saw the additional options "magically" appear on an Abit AN8 Fatality motherboard when a 4200+ x2 processor was used.
Unfortunately the long-rumored 1T Command Rate with Rev. E AMD processors appears to have been just a rumor. We could not run 4 matched dimms at 1T in a motherboard with Rev. E/x2 support. Four dimms still required a 2T Command Rate.
You should consider our list of Official and Unofficial Dual-core support to be a unified Revision E list. AMD tells us that Revision E support is part of the requirement for x2, so you can assume any x2 capable board will also properly support Venice and San Diego.
Fortunately for those shopping for a new Socket 939 motherboard there are many new boards that support Revision E and x2 processors. We hope the information in the article will make your search for a new motherboard for an Athlon 64 an easier process. If we overlooked boards and/or BIOS revisions that belong in our Supported list please let us know.
Facebook
Twitter
RSS
57 Comments
View All Comments
PrinceGaz - Wednesday, June 29, 2005 - link
#32- those TWINX memory products are matched pairs of 1GB modules, which gives a total of 2GB. They are not individual 2GB modules. AFAIK all of the 2GB modules available are registered ECC such as Crucial's 2GB PC3200 module CT25672Y40B. That makes them unsuitable for A64's, but fine for Opterons (which is what you would expect as modules of that size are only likely to be used in servers).I'm almost certain that A64 Rev.E DOES support four DIMMs at 1T command rate, but ONLY if single-sided. Being able to use 1T command-rate is dependent mainly on the load placed on the memory bus, ie the number of memory chips on the channel. Two double-sided modules per channel is 32 chips which is never going to run at 1T.
#27- all 1GB modules are double-sided, it's highly unlikely that your freind with four 1GB modules is running at 1T, or if he is that they are running error-free. I suggest he lets Memtest86 loop through all its tests overnight as it will probably report errors (test 5 in particular).
smn198 - Wednesday, June 29, 2005 - link
#28 Wesley - I also would love to see the performance of dual core with the new async. memory speeds. I imagine that dual core would benefit dual core more than the FX57 but I'm not sure how bandwidth starved a dual channel X2 is."AMD has also added additional "hidden" features in the AMD on-processor memory controller. Additional asynchronous ratios are available at 433, 466, and 500 memory speed on boards that implement the necessary code to access these memory controller features. These options should be available with any Revision E chip if the manufacturer implements the controller option read in BIOS."
sprockkets - Wednesday, June 29, 2005 - link
anyone want to make a wiki for processor support here for AMD stuff?justly - Wednesday, June 29, 2005 - link
"If we overlooked boards and/or BIOS revisions that belong in our Supported list please let us know."The Asrock K8Upgrade and Combo-Z also have X2 BIOS support.
Slaimus - Wednesday, June 29, 2005 - link
What about single sided sticks? I have two single sided 512MB sticks in my current system specifically to allow two more to be added.jiulemoigt - Tuesday, June 28, 2005 - link
Your not actully saying the same thing 4 gigs ram != 4 sticks of ram, there are 2gb sticks just not many:TWINX2048-3200PRO 2048MB 3-3-3-8 2x184
TWINX2048-3200C2 2048MB 2-3-3-6* 2x184
TWINX2048-3200C2PT 2048MB 2-3-3-6* 2x184
TWINX2048-3200 2048MB 3-3-3-8 2x184
TWINX2048-3200PT 2048MB 3-3-3-8 2x184
gettting 1T commands out of any of those sticks on the other hand may simply not be possible, the problem was not the size of the memory but shared latency, with four sticks you have four over lapping fields that have to line up really nice. With 2gb sticks you only have two fields :)
cryptonomicon - Tuesday, June 28, 2005 - link
"Unfortunately the long-rumored 1T Command Rate with Rev. E AMD processors appears to have been just a rumor. We could not run 4 matched dimms at 1T in a motherboard with Rev. E/x2 support. Four dimms still required a 2T Command Rate."Doesn't surprise me... thats the price you pay for the amd platform. However its pretty much the only bad part.
Wesley Fink - Tuesday, June 28, 2005 - link
#27 - People who know memory also tell me no 2T with 4 dimms, so I am skeptical of your friend's claim. Sandra appears to report 1T when it is not really reporting Command Rate which confuses many. I know Oskar's BIOS' for every memory - I believe last count was more than 55 BIOS revisions for the DFI nF4. That's not the issue. The 4 dimms 1T would be a function of the mem controller on the new Rev E and not have a lot to do with the board. The new Rev E mem controller is indeed a better overclocker than the earlier 0n-CPU controllers.KeithDust2000 - Tuesday, June 28, 2005 - link
Wesley, thanks a lot for the informative and prompt reply, I can´t wait to see the results!:)
Wesley Fink - Tuesday, June 28, 2005 - link
KeithDust2000 - Wish I had an answer to that. I'm still waiting for a FX57 and/or 4800+ from AMD for our motherboard reviews. When supplies ease a bit and I get the CPUs I'll include some asynch benchmarks in a future review. The difference will not be dramatic, but with DDR500 available at pretty fast timings these days, you will get a bit of a boost from a 400 CPU running at stock and driving memory at DDR500 or 533.Only a few boards implement the new memory options correctly, and that may be why a lot of reviewers overlooked them. AMD didn't advertise the new ratios in their press kit and most weren't looking for them.