The Inquirer-Home

Asus pumps up Nvidia's Nforce 790i reference design

First INQpressions Asus Striker II Extreme mainboard plus twin 9800GX2 in QuadSLI
Mon Mar 31 2008, 15:47

Product Asus Striker II Extreme mainboard
Chipset Nforce 790i Ultra
Website www.asus.com


JUST OVER A YEAR ago, Asus' Striker Extreme, based on the Nforce 680i - but not the Nvidia reference design, coupled with quad-core Xeon flavour of Core 2, that was our PC benchmark winner here. A combination of decent CPU overclocking plus very high FSB 1667 and in-sync low-latency DDR2-833 CL3 still matches even the newest Penryn-based systems pretty well. It did have occassional stability issues, but the newer BIOSes numbered 1300 and later seem to have solved most headaches.

In the meantime, Nvidia slowed down on everything, chipsets included. Only this month saw the true successor to the 680i - the Nforce 790i Ultra (yes, there is a non-Ultra 790i as well, with lesser memory overclocks). True PCI-E 2.0 slots, new DDR3 memory controller and a bit cooler operation, plus a new stepping of the old 570i South Bridge - that's about it.

Again, instead of re-badging the Nvidia reference mobo, Asus did its own design - here we look at the Striker II Extreme, the 790i Ultra entry. And yes, we combined it with two Asus 9800GX2 cards for a QuadSLI setup, too. Here are the first quickie impressions - more after the Iintel's Shanghai IDF business.

As with all Republic Of Gamers (ROG) boards, this one comes in a fancy box with a myriad accessories. For instance: the LCD status monitor - very useful for diagnostics, as well as VRM cooling fan - not so useful, unless you double its rotational speed. The board's trademark ROG copper heat pipe system is simpler, compared to the three-storey maze seen in the old Striker Extreme.

This one is far more advanced: the "Fusion" North Bridge cooling portion incorporates a combination of an integrated waterblock and a large heat sink array. The ultimate option is using that watercooling and still ensuring good airflow about the heat sink portion - important if trying to do FSB2000 on a quad-core CPU with three FSB loads. Without water cooling, the NB quickly exceeded the 50C temperature threshold at this FSB setting, and even the Sandra tests couldn't complete.

Like all other Nforce 790i mobos, Striker II Extreme has three full x16 PCI-E slots: two blue at v2 speed (2 x 8 GB/s per slot) coming from the North Bridge, and one middle white v1 speed (2 x 4 GB/s per slot) from the South Bridge side. For the standard two-card SLI, please stick with using the blue slots only: not only is the slot transfer speed doubled (less critical) but, when communicating over PCI-E, the two cards need not make further latency-intensive hops over HyperTransport between North and South bridges.

Talking about that communication, Striker implements 790i's multi-GP broadcast and direct PCI-E GPU-to-GPU talk options without going through the FSB and memory again. All that pending driver support, of course.

Besides usual graphics, an even more critical, but still seemingly overlooked, benefit of these features would be in Linux-dominated GPGPU computation applications, assuming Nvidia implements them in its Linux drivers. These latency-reducing and bandwidth-saving features could speed up many routines that occassionally send and receive stuff from the main memory - those that do that main memory shuffle non-stop since graphics memory size limits aren't cut yet for GPGPUs - until we see Larrabee sitting pretty directly on QPI next year.

The board supports up to 8GB DDR3 memory, including the new EPP 2.0 spec, a direct competitior to Intel's XMP - so, there really is a need now for DIMMs incorporating three settings: basic, XMP and EPP2. Unless, of course, Intel and Nvidia somehow agree to merge the specs which, in their present relationship status, is less likely than the sight of natural snow falling on the pretty streets outside my Singapore home.

Other I/O stuff is the same as on the previous 680i and 780i chipsets: no surprise since the South Bridge remains the same. The software-driven sound is, thankfully, on a daughtercard, without blocking the valuable mobo real estate. The MediaShield SATA Raid and intelligent, accelerated twin Gigabit Ethernet ports with teaming are one plus here over the Intel chipsets, aside from supposedly still faster USB2 real bandwidth. Again, I commented to Asus repeatedly to try to keep both the PS/2 keyboard and mouse ports, but to no avail.

The Award BIOS used still sticks like a sore thumb compared to the smooth BIOS setup menus of Asus Maximus Extreme X38 chipsets. The overclocking options are still somewhat wider than on Nvidia's reference design, though.

The setup configuration, on which we'll be running the benchmarks after IDF is: Striker II Extreme with BIOS 0402 or newer; Intel QX9650 CPU, with extra runs on QX9770 and Xeon 3360 45 nm offerings, as well as the QX6850 "old" 65 nm top-notch Kentsfield. As for the coolers, I ran Thermaltake DuOrb, but had to give up the standard mounting orientation as it conflicts with the North Bridge fins. The benchmarks are all run on the Corsair Nautilus 500, and we'll be trying some other interesting stuff too, including water cooling the chipset in the second run.

alt='striker2'

As for memory, we didn't have anything EPP2-ready on hand, so we tried something special: OCZ's brand new 2 x 2 GB heat-piped DDR3-1333 CL5 DIMM set. It can't run 1:1 on high FSB settings (yet, this board has 2-phase DDR3 power, so there is potential), but the capacity is there to match Skulltrail in all Vista64 tests.

The graphics is matching: twin Asus 9800GX2 dual-GPU cards for Quad SLI - these already ran in four-GPU mode effortlessly on the Skulltrail. True to that, the Quad SLI ran fine in the default test runs before overclocking - see the results below.

alt='3dmarkstriker2-9800gx2def'

DEFAULT

alt='3dmarkstriker2-9800gx2slidef'

DEFAULT SLI

alt='3dmarkstriker2-9800gx2wuxga'

WUXGA 1920 x 1200

alt='3dmarkstriker2-9800gx2sliwuxga'

WUXGA SLI

I also did some initial overclock fiddling: the in-build "Crazy" CPU Level Up mode doesn't exactly push the CPU to the limits. It pushed the 3 GHz QX9650 only to 3.5 GHz - not much for a CPU that runs fine up to 4.3 GHz within safe settings on standard watercooling. The board and the CPU can do far more when driven "manually".

alt='strikerbios'
Once you're in Windows, there are software tools like AI Gear 3+ with AI Overclocking (kind of intelligent CPU frequency tuner) and ASUS AI Booster Utility. Just in case you go overboard, Asus still has COP EX (Component Overheat Protection) and aptly named C.P.R.(CPU Parameter Recall) with the usual onboard switches for power, reset and clear CMOS. Keep in mind that, in our experience, a few of these tools - even Asus' standard PC Probe2 - do crash occassionally in Windoze Vista64 flavour. So, the old XP may yet be a more stable OC platform.

Generally, we found the first run OK. We managed to boot the QX9650 at 4 GHz / FSB2000 with memory stuck at DDR3-1333 and no, it didn't complete the tests with the initial voltage settings we played with: NB watercooling will help there. The 4GHz / FSB1600 setting ran fine after half an hour of voltage tweaks but no, we haven't run Linpack or PCmark Vantage to confirm true stability (which we have on the X38/X48 by the way).

In summary, this is the top end of Nvidia-based mobos. Foxconn, MSI and Gigabyte, as well as EVGA, all have high-end 790i Ultra boards as well, and the comparisons will be interesting - especially with the ultra-tuned, water cooled Foxconn Quantum Force entry.

Within the Asus' own ROG line, the direct competitors will be the upcoming Rampage Extreme board, as well as, quite possibly, the "market test" prototype called Pinot Noir: the sexy red-black, fully watercooled, X48 ROG mobo that Asus showed at CeBIT. According to Asus managers, it was there to gauge the market feedback and we hear the feedback was good.

Look ahead to more tests on this mobo in the next few days. ?

 

Share this:

blog comments powered by Disqus
Advertisement
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

Advertisement
INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?