EPoX EP-8NPA SLI and the A64 4000+ mobile processor
EPoX claims that the EP-8NPA SLI is an ethusiast's motherboard for owners of Socket 754 processors. I should have noticed that their website stressed that you didn't need a fast processor in order to maintain great SLi based performance.
My personal system was based on a Chaintech VNF3-250 motherboard with a fully unlocked 6800LE videocard (leadtek) and depending on what settings I was using in BIOS I used HyperX memory of either DDR33, DDR400, or DDR500 speed (I have at least one 512MB strip of each). Installed in this system was an Athlon 64 2800+ mobile processor (which worked perfectly). In combination this system flew, but the mobile 2800+ is clocked slightly slower than the desktop version. It overclocks well but runs surprising warm (even with good air cooling).
To make a long story shorter, I purchased the first A64 4000+ mobile processor I could get my hands on. This processor involves a die shrink, it run cooler, has SSE3 instructions, and an improved memory controller. AND it works flawlessly in the Chaintech VNF2-250.
When combined with the parts from above system, what flew before blasted past the sound barrier a when the 4000+ mobile processor was added. Adding an Audigy sound card further spped up the system to the point where my benchmarks were significantly hirger than many Socket 939 system (and ALL socket 754 systems).
Then I decided I' d like to try a PCI-e graphics card. While there were many Socket 754 Motherboard on the market that would have worked with my existing components (except, of course, for the video card) I chose the EPoX EP-8NPA SLI because in addition to PCI-e it offers SLi support, SATA2capability, and gigabyte ethernet networking. Sounded like heaven.
WELL, next time I'll do a little more reasearch. The EPox motherboard DOES NOT support mobile processors in any fashion. My 2800+ and 4000+ processor REFUSE to post on this board. EPoX techical support claims they will not be adding this feature to future BIOS revisions (as they state that there are significant differences btween the DTR line and the standard desktop line).
As we all know, currently there are only minor diiferences between the DTR's and the current desktop units - usually requiring a few minor BIOS tweaks to recognize these units (Chaintech has had NO problem supporting DTR processors - in fact I didn't even need a BIOS upgrade to move from the 2800+ processor to the 4000+).
I had resort to obtaining an A64 3000+ desktop processor (Newark core - surprisingly similar to to core in the new DTRs) to test the EPoX MB for functionality. It works fine, good overclocking feature in BIOS, and a nice combination of harware features (however, obviously, it won't rcognize the DTRs). In addition, it soon became apparent that the fastest processor that this MB can support is a 3700+ desktop processor.
This is a shame for several reasons. First, I already have the DTR 4000+. Second, EPoX is not supporting the fastest processor their MB COULD support (with minor BIOS revisions). Third, I already had the 4000+ clocked up to 2880 GHz aircooled with only an minor upgrade to the processor cooling fan/heatsink. I don't think (without exotic cooling) that the 3700+ is nearly a cabable an overclocker. In fact, I'msure that in the long run I could have gotten even more performance out of the 4000+ (I have watercooling equipment I have as yet to try). Finally, the whole sistuation places EPoX in a less competitive position than their rivals. While I would like to retain the features offered by the EP-8NPA SLI I intend to continue using the mobile 4000+ (which probably means returning it to the Chaintech MB).
Pity, EPox may change its mind (on supporting the DTR), but I doubt it. A third party may be able to revise their BIOS (and that is my next area of exploration). Or, I may just wait for the Socket M2 to perform another upgrade. Only, time will tell.
IN THE MEANWHILE, IT IS IMPORTANT FOR T.H. READERS TO BE AWARE THAT THIS MOTHERBOARD IS COMPLETELY INCOMPATIBLE WITH DTR PROCESSORS (and that there are plenty of alternatives to choose from that will work fine with said processors).
I'd like apologize for the overly long post and I wish you all well in this matter.
Tech support proved wrong again!
The same trick I initially used to run Athlon XP-M processors in my Asrock K7S8XE+ motherboard works with the EPoX EP-8NPA.
Had standard A64 3000+ processor installed in this board (just to get it up and running). I then went into BIOS and disabled Cool 'n Quiet and set the processor multiplier to its lowest setting (4X).
I then reinstalled the A64 4000+ mobile processor into the EPoX EP-8NPA.
Initially, I booted the system at standard voltage (because EPoX limits processor voltage selections to standard, plus .25 or plus .5v, and a wide selection of minus/undervolted settings).
And, lo and behold, the system booted at 800 MHz.
First success logged.
Next, obviously as EPoX's problems stems from how the BIOS controls voltage selection we need a tool to work around that problem. Enter an old software hack CPUMSR. This program allows processor voltage and multiplier changes on the fly within Windows. Once the system was up and running Windows (@ 800 MHz) I used CPUMSR to change the processor voltage to 1.4v and I increased the multiplier value. There are a few rough spots that I'll discuss in a minute, but IT WORKS!
Now, on to the the rough spots. First, for some reason, once I used CPUMSR (to alter processor voltage) the system boots at 1.4v (I'm not complaining, its just a little weird). Next, I could not get CPUMSR to adjust to processor frequency above 2400 MHz. This is an obvious 12X multiplier (not the maximun 13X shown by the program).
Oh well, its a start.
AND I PROMISE YOU I WILL GO FURTHER!
F'ing A! This is too cool!
Thanks for the support and encouragement.
Bear, Delaware, USA
Since the system now boots at a processor voltage of 1.4 (or higher if I set the BIOS to plus .25 or .5v) the next step was obvious. Enter BIOS and make processor multiplier and FSB selections directly. First, test at a stock speed of 2600 MHz (13x 200FSB) - no problem.
Now, with a 12x processor multiplier times a 240 FSB (with memory set at DDR333 to utilize my low latency DDR400 memory) I am now running at 2880MHz (again)!
BTW, thanks for the encouragement. It's difficult to believe I only got one positive response (and that was from the other side of the planet!).