Maxing out

The final challenge was to strap a cascade to the CPU and see how far we could push things with the additional overhead sub-zero temps can provide. Those of you into this sort of thing will be happy to know that we found no cold bug bootup limits on the board itself. Nehalem processors are known to average bootup limits in the region of 50-80C. Our retail i920 processors (purchased from Scan UK and Tank Guys) will boot on this board as cold as our smaller cascade will go, which is -95C (sounds lucky at this point does'nt it?).

There is one huge caveat, though: although our i920 CPU boots all the way down to the limits of the cascade, overclocking it at sub-zero temps is another matter altogether. Anything over a reference BCLK of 166MHz on the 20X multiplier refuses to boot. We added a number of shims between the evaporator and CPU to increase the temperature delta to find where this situation begins to manifest. Our saved 195 BCLK x20 multiplier BIOS profile loads fine as long as the CPU is in the positive temperature region. Anything on the negative side and we're stuck at near board defaults for voltages.

The major obstacle for us was that we were limited to using "BIOS-boot only" overclocking. As we mentioned on page 2, the AEGIS panel and the last version of SetFSB did not work properly for us. Had AEGIS panel worked, we would have been able to tinker with voltages at the OS level perhaps providing us with enough room to manipulate the BCLK reference clock to favorable levels. BIOS voltages for the CPU and VTT/Uncore were raised and lowered in BIOS to see if we could get around any kind of current limiting at negative DTS readings; unfortunately, these attempts proved unsuccessful. What we ended up with as workable voltage put us in a worse position than when we were using water-cooling at ambient temps.

It seems we have an i920 processor with an odd cold bug, or perhaps we're seeing what could be some kind of power throttling due to a built-in CPU thermal condition breach. Nehalem monitors processor Vcore at all times and can make dynamic changes on-the-fly in an attempt to rectify temperatures that fall above or below predefined DTS thresholds. Still, this is another story altogether and not something we're blaming the board for just yet. Although we are beginning to see BIOS's on other boards that offer partial workarounds to some of these conditions. A hard modification will be required to the Blood Rage to overcome some fo the current throttling at the PWM end of things, and this is an avenue we will explore. Before doing that, we're going to try out a few more processors to double-check everything and report back in the full review.

After all that effort, it was back to water-cooling the processor, and we managed to get a clean boot at 200 BCLK on the 20-21X multiplier and run 3DMark06 along with several other benchmark programs and applications.



We did manage to get the board to boot at 203 BCLK, but it seems our processor is too VTT hungry (at least on this current BIOS) to scale past this at voltages we'd deem being past intelligent for ambient benchmarking. The i920 series is hampered in this regard by Uncore and QPI MHz limits, although we did try lowering the multiplier/memory speed to work around this without success. Future tests will be compared against a retail Extreme i7-965 CPU to get a broader picture of motherboard capabilities.

It's All About Brawn... Well, Maybe First Impressions
Comments Locked

66 Comments

View All Comments

  • davekozy - Saturday, January 3, 2009 - link

    Asus is already releasing version 2 of the P6T Deluxe less than 2 months after the first was released. From what I've read the main difference is the loss of the SAS connectors. Companies with good product support release bios fixes as often as needed. There are 4 bios updates for the first P6T Deluxe or about one every two weeks since release. There have been 8 so far for my 6+ month old Asus P5Q. Evga has 7 bios updates already for their X58 board. That's almost one a week! There are a lot of issues the beta-testers aka early adopters (including AT) have helped resolve.
  • gemsurf - Saturday, January 3, 2009 - link

    Sorry if I let the dogs out Rajinder and I hope it is all taken in context. We don't have the voice that this site does and thru you maybe some of this can improve? We have all trusted Anands for a very long time and I certainly hope he is reading these comments as well. Thank You all for what you do!
  • Rajinder Gill - Saturday, January 3, 2009 - link

    Not at all. The reason I did not sensationalize the issue is becasue many others are not experiencing the problems. It's still random. I plugged in an OCZ Pro Xstream today and the board works fine. The variables involved in parts mismatches, batch changes, underhood revisions are so numerous. In the end I felt it better to highlight the possibility. Might not have been what you wanted to see, but the point was there regardless, and from my perspective understood too.

    later
    Raja
  • Scott66 - Sunday, January 4, 2009 - link

    Another website had to switch to a different power supply as their usual benchmarking one did not cold boot the system.
  • kilkennycat - Friday, January 2, 2009 - link

    Have you noted that there is a recession? Core i7 is nowhere near mainstream, the current version of Core i7 is not optimised for desktop, being too power-hungry, and the motherboards and memory are ridiculously-priced. Maybe I am "all wet", but it might be very revealing to generate a survey of Anandtech readers with regard to their computer-system purchase intentions within the next 3-6 months.
    Why 3-6 months? The current bunch of Core i7 motherboards are very obviously first-generation beta-phase efforts. It takes somebody with a whole bunch of money and an "early-adopter" mentality to have the fortitude to buy any of them. It is going to take 3-6 months for Core i7 to be considered mainstream and by that time Anandtech is going to have to update a whole bunch of these X58 motherboard reviews. Meanwhile, there is a huge price/performance sweet-spot in the Penryn and Phenom domains with some incredibly comprehensive offerings out there. For just one little example, it is disappointing that the full Anandtech review of the MCP7A/9300/9400 motherboards has continued to be delayed, presumably by the pressure of completing these "bleeding-edge" X58 reviews.
  • ssj4Gogeta - Saturday, January 3, 2009 - link

    I'm probably going to wait until Westmere.
  • LoneWolf15 - Friday, January 2, 2009 - link

    Yep.

    After having to purchase a board unexpectedly this week, and some thought, I've decided on Gigabyte's GA-EP45-UD3P.

    Current reviews almost seem to make it look dated, despite it being a relatively recent board. But considering I'm going from an Intel BadAxe2 (when I buy a board, I want a good board that will last a few years), a P45 chipset is quite new. System also supports DDR2; quality DDR2 is still half the price of DDR3, and it supports faster processors than I need (including my existing Q6600). Add that it also has ICH10R for RAID, firewire, and a ton of ports, and I can save a lot of money over getting an i7 that is certainly cool, but beyond what I need. All for under $150 shipped.
  • Paulywogstew - Friday, January 2, 2009 - link

    In the Board & Features overview you state its a socket T 775 interface???
  • Kroneborge - Friday, January 2, 2009 - link

    IMO, bios revisions for better performance etc are ok, and to be exptected. Bios revisions to get basic functionality working are not.

    When you make a purchase you expect it to work as advertised or it's going back. If you can't get a bios ready by then, then delay release. Develop a repuataion for rock steady performance, and people will beat a path to your door to develop your product, EVEN if they have to wait a extra month or two too purchase.
  • JonnyDough - Friday, January 2, 2009 - link

    Agreed. Every time I took that brief AMD survey upon a visit to their site I always put that I looked for STABILITY in drivers. Why would I want to get 300fps in a game if it's going to crash on the last level and not let me win?

Log in

Don't have an account? Sign up now