GLEAMING THE STRIKER CUBE . . .

BonzaiDuck

Lifer
Jun 30, 2004
16,303
1,872
126
Over the last year and a half, I've been salting my posts here with statements that "A major C2D build" was in the works, with vague allusions to a patient review-search and analysis before I put down my money.

I wanted originally to use the much-touted P5WDG2-WS-[whew!]-Professional motherboard -- using a 975X chipset. It was a proven choice -- a bit expensive for many, a 5-out-of-5 stellar review by perhaps 20 customers at NewEgg, posts by builders in threads at various geek web-sites since last August or September. It would allow me to implement hardware RAID-5 using a 128 or 256 MB on-board cache, fitting a PCI-X slot on the motherboard. I expected to get a sustained throughput of up to 200 MB/sec and certainly double the sustained throughput of a single Seagate "perpendicular" or WD Raptor drive. But the card worked in half-duplex mode through what is otherwise a very adequate PCI-X bus speed -- far in excess of a drive array's limitations with either of those two drive choices.

Alas, the 680i boards seemed "risky." I know some "closet noobs," and they were acting like chicken-little per problems some people have had with one or more of these boards, the chipset -- but mostly the BIOS upgrades, since at least one board I know of has gone through some four or five BIOS revisions since last August.

But I wanted those 46 PCI-E lanes. I wanted full-duplex advantages and an extra slot guaranteed to run x4 and x8 controller cards, while still running two Express graphics cards in SLI with both working concurrently at x16. I couldn't have the SLI "upgrade" options with 2 x16 plus 1 x8 [x16 slot] in the WS-Pro board.

I figured it was time to take a chance. I veered from a preference for Intel chipsets and bought the Striker Extreme board. I plopped down some $370 for the board in March, to see the price drop again to around $335 to $350.

The other things I hoped for with that board included 8-cycle VR (as with the "WS-Pro" board): "Stack-Cool" technology, the potential for good ducted cooling with built-in copper heatpipe cooling for SB, NB and VR components. I wouldn't have to "sink" my Mosfets.

With a test-case that is just a temporary home for the basic components, I have yet to install a PCI capture-tuner card and the PCI-E x8 Areca 1210 RAID controller. I might have my eyes on the 1220, but RAID6 needs at least four drives, and I intend to start with three. I chose the Seagate 7200.10 320GB "perpendicular" drive to compromise 4 MB/s in sustained throughput for double the storage of a 150GB Raptor drive.

EXPLODING MYTHS ABOUT THE STRIKER BOARD

A customer review in a tone of knowledgeability says "You won't ever get beyond 1,333 Ghz with this board." Not true. First, it depends on the processor, and second, it depends on how much you want to over-volt that processor. With much greater confidence than I have for customer-reviews, CDInfo and other sites show the board to be immensely over-clockable beyond 1,333 Mhz FSB.

"I can't get it to boot under any setting." Another complaint -- it appears to be a confusion about the "On-Board LED" switch on the board's I/O panel. The switch uses the same circuit applied to the board's power-switch and the three on-board [LED-green-red-green] switches -- "Reset," "CLR-CMOS," and "PWR-ON." Pushing the "OnBoard LED" button lights either the "GRG" switches or a set of blue-LEDs -- but not both (although there may be a non-default setting in the motherboard that would allow it.) The Blue LEDs supposedly assist in finding various pin-outs and plugs on the motherboard.

But you cannot boot the board without turning off the blue-lights! So despite the general -- almost inscrutablely brief nature of ASUS manuals, the warning note is "right there." Gamers seeking fancy motherboards would do well to read the manual.

"I can't get it to work with my Corsair memory." Some Corsair memory apparently doesn't configure well to the Striker, but other Corsair memory is proven for it. Gamers seeking fancy motherboards would do well to search compatibility lists at Corsair, OCZ, A-Data, Crucial, Patriot, Mushkin and other web-sites.

CUSTOMER REVIEWS, MARKETING AND MARKET SHARES

By appearances, many of the consumer-reviews posting 5-stars for the WS-Pro board seemed more "knowledgeable." They were almost exclusively people who wanted the PCI-X slots and knew something about "hardware RAID." But what could you expect for Striker customers, when the board has almost an adolescent theme-name appeal? Frankly, "Striker Extreme" and "Commando" put me off, when I'd been used to a naming convention similar to "P5N32-E SLI" or "P5W64-WS." Or even "P5WDG2-WS-[whew!] . . Pro."

Whether the board is a "gamer's" board or a high-end "workstation" board, it is a "Cadillac" among boards. A customer reviewer said something like "This is a GAMER'S board, Pindaho! If you want a SERVER board, get a SERVER board." But there isn't any hard and fast line between these products in either the "Republic of Gamers" or the "Workstation" category. I'll buy the features I need or want.

CONTAIN YOUR RISK

The Striker and some of its cousins are spec'd to run above the current "mainstream" FSB speed of 1,066, and ASUS promises 1,333. They certainly -- and easily -- deliver. But here are better reasons why this might be significant.

The board is warrantied to run at 1,333 Mhz FSB. But there aren't any extant and reasonably-priced processors that run at that speed, if at all for the moment. Running the board above 1,333 may technically void the warranty, but even 1,600 Mhz represents only a 20%-above the spec. You may void your warranty, but you would probably be in little danger of needing it in the first place.

The board runs in "native-DDR2-800" mode. If you purchase the right high-performance dual-channel kit, you could run the board up to 400 Mhz [external frequency] and be in no danger of voiding the warranty for another part.

That leaves the CPU. The Striker seems an open invitation to some noobie running up the VCORE and turning their CPU into a crispy-critter, because it is adjustable to 1.9V VCORE. Similarly, the vDIMM memory voltage allows for settings up to 3.5V-something -- way beyond the warranty limit on most high-performance DDR2 modules of about 2.2V.

EASY "Auto" CONFIGURATIONS AND EASY OVER-CLOCKS

You can drop in a C2D processor spec'd for 1066 FSB, such as the E6600, add memory like the Crucial Ballistix DDR2-1000's, and the board will configure to run the memory at DDR2-800 with an automatically chosen divider at the processor's stock speed. Easy. No instability. Runs very cool.

I want to speculate about the customer who said the board wouldn't run above 1,333 Mhz FSB and other complaints. The "FSB and memory" settings provide a "linked" option allowing manual selection of FSB speed with memory running at a "linked" speed determined by a "ratio" setting in that menu. I can imagine people selecting an FSB speed and an FSB : DDR ratio of 1:1. But even the ASUS manual denotes FSB as "QDR" and memory as "DDR," so the proverbial "1:1" ratio users should seek under this menu's "linked" option would be 2:1 in order to achieve the desired 1:1 ratio between FSB and dual-channel memory. Choosing even a speed of 1066 FSB under "linked" with a 1:1 menu choice would run the memory at DDR2-1066. Under that sort of confusion, anyone would say "you'll never get it to run above 1,333."

CDInfo's review of the Striker board showed in detail how to run the FSB @ 1600, the memory at DDR2-800 so that all other settings -- particularly the voltages -- are on the default "Auto" setting. To do this, one needs to drop the multiplier from 9 to 6, and an E6600 processor would be running at 400 Mhz above stock, or around 2.8 Ghz-something.

But there are "sweet-spots" for multiplier settings of 7, 8 and 9. And getting to those sweet-spots depends on manually setting the VCORE and other voltages.

CORE-2-DUO VOLTAGE LIMITS

On another forum at www.i4memory.com, a poster under the name "Eva2000" ran some very methodical tests with the Striker and an E6600 processor. She (I assume from the prefix "Eva") -- set the VCORE to 1.525V, the VDIMM at 2.1V, the 1.2VHT to 1.4V. This would allow running the 2.4 Ghz E6600 @ 3.6 Ghz on air-cooling, with the memory running at DDR2-800 in a 1:1 ratio. There should be some data from her posts about heat.

According to Joe Citarella ato www.overclockers.com, thermal power increases linearly with speed in megahertz, and it increases by the square of the VCORE value. I discovered that temperatures under ORTHOS stress-testing would increase by as much as 25C-degrees over idle at VCORE in excess of about 1.475V, and this would lead to instability under load that might not be detectable without the stress-testing. Correction would lead to further voltage increases and possibly higher load temperatures.

This seems to have little to do with the motherboard. With E6300 or E6400 processors, no VCORE adjustment -- or very little -- is required to push the processor to a 50% over-clock. But that's the way it is in a processor-line like the C2D.

I chose to do certain things per VCORE regarding my over-clock attempts.

First, find "candidate" setting-ranges for speed at the multiplier settings of 7, 8, and 9. Second, arbitrarily limit the over-volt settings to a common value in determining these settings. Third, keep the VDIMM just above its operating range median, or just below 2.1V. In this case, I set the VDIMM value to 2.05V, with the monitored value shown to be around 2.09V.

I finally chose a VCORE over-volt setting of 1.4625V. If the Intel spec is 1.35V and refers to load voltages (as opposed to idle values), this means that the setting exceeds the spec by 0.07V or around 5%, because the voltage-droop I experienced was about 0.035V, putting the monitored reading at about 1.42+V. Obviously, the amount of voltage droop might be a reflection of how well the motherboard regulates voltage -- with "capless design," "Stack-Cool," 8-cycle VR, and the copper-heatpipe "necklace" that covers the VR Mosfets. Some people reported a voltage-droop of maybe .01 to .015V with E6300 processors under "auto" settings. Voltage droop might be proportional to the absolute voltage setting, but also related to the motherboard's voltage regulation design.

This generated load temperatures in a 43C to 47C range during Orthos testing. Those temperatures can be further reduced by either effective motherboard ducting or water-cooling. And in milder stress tests, the idle-to-load spread is within -- or close to -- about 10C-degrees -- if the cooling apparatus has an effective thermal resistance of 0.12 C/W or less.

BIOS INACCURACIES

The three voltage rails were tested by "Eva" with a multitester to show that BIOS and windows monitored values are biased downwards. This confirmed for me that my rail readings pertaining to my Seasonic 650HT PSU were understated, and that the Seasonic is about as close as it gets to maintaining the rail requirements. The understated voltages were off by some 3% across the board (a pun, ha!) But there were other indications from forum posts that the "set-voltages" and actual voltages differed.

These experiences occurred with versions 301 through 701 of the Striker BIOS. The board has since been through three more BIOS revisions, and with version 1004, it seems likely that the set-values and "real" voltages are consistent, while the monitored values differ from either the set or "real" values.

STUNNING RESULTS FOR SOMEONE NEW TO c2d, DDR2, AND FASTER FSB SPEEDS

For E6600, Multipler = 7, VCORE= 1.4625V, VDIMM 2.05V:

About 2.8+ GHz CPU, FSB = 1600, DDR2 = 800, external frequency 400 Mhz

The FSB and memory can be pushed further here -- I ran it up to about 1616 FSB and DDR2-808 so that I could find a more comfortable "wiggle-room" to drop it back to 1600.

E6600 -- Multiplier 8, same VCORE and VDIMM:

384 Mhz external frequency translates to DDR2-768, FSB 1536, CPU = 3.07 Ghz

[More stability testing needs to be done for this multiplier choice, but adjustment to the frequencies should mean an external frequency somewhere between 375 and 384.]

E6600 -- Multiplier 9, same VCORE and VDIMM:

360 Mhz external frequency, DDR2-720, FSB 1440, CPU = 3.214 Ghz


WHAT MEMORY LATENCIES?

This all depends on the selection of dual-channel kits. The Crucial Ballistix DDR2-1000 modules will run at DDR2-800 speeds with latency settings of 4,4,3,8,[1T or 2T], and those latencies should be good between memory speeds of 750 to 850 Mhz.

I have yet to try tightening the latencies to achieve "CAS 3" at lower over-clock settings, beginning with CPU = 3.0 Ghz, FSB = 1334, DDR2 = 668.

Other memory kits by OCZ, Mushkin, Patriot etc. should also work at tighter latencies at these speeds. The DDR2-1000 modules have recommended latencies of 5,5,5,15 at their rated speed, but those values can be reduced to 4,4,4,12, or 4,4,4,10, or 3,4,4,10 -- but especially to 4,4,3,8 if run at around DDR2-800.

I don't care what I've heard from others -- especially in reseller reviews. The improved latency timings provide better bandwidth at the speeds I've stated. You can also run the Crucials at a 1T command-rate, and that choice, too, seems very stable and reliable.

TEST-SOFTWARE SUPPORT IS LAGGING . . .

SpeedFan incorrectly reports the nVidia 8800 GTS temperature as a "CPU COre 2" temperature.

ORTHOS, SnM, PassMark CPU-Burn-in-Test incorrectly report CPU speed as a product of the chosen external frequency and the stock multiplier of 9, even with lower choices for the multiplier.

nVidia Monitor correctly reports the CPU and other speeds. Voltage "reports" by nVidia Monitor are "set-value" voltages -- not monitored voltages. If I didn't already say so, the "set-value" for the Striker's default-"Auto" VCORE is 1.44V (idle). Applying my own observation about voltage droop with the E6600, this means that the load VCORE would be somewhere around 1.40V -- and 0.05V ABOVE the Intel spec.

GLEAMING THE STRIKER CUBE

$350. That's 350 clams . . . 3.5 Franklins. About double the price of several good motherboard choices. for instance, the price for a P5B (965 chipset) motherboard is just over half the reseller price for the STriker.

If you want the enhanced voltage regulation features with a 680i chipset, you need to decide if it's worth an extra $150 to get it. The cooling enhancements to the NB, SB and VR Mosfets would eliminate a lot of inconvenience playing with thermal epoxy, aluminum Mosfet and PLL sinks, and voiding your motherboard warranty. I have yet to replace the thermal-paste on the copper-heatpipe "necklace" with Arctic Silver 5, but I would expect a slight improvement in temperatures.

For the price, the board offers an immensely adjustable BIOS setup, the voltage regulation features, the 680i chipset, three PCI-E slots to run two graphics SLI cards each at x16 with the third slot providing an opportunity to run another expansion card at x8 without degrading the graphics.

Some users -- several, in fact -- have posted observations that the board either conflicts with, or generates noise for -- Creative Labs X-Fi PCI sound cards. The trouble seems to be noise generated by the PCI-E bus. To resolve that problem, ASUS moved some of their onboard "SoundMax" audio to an audio "riser" card that fits in a proprietary slot. The slot can be easily confused with a PCI-E x1 slot. Some two years ago, Creative Labs and SoundMax collaborated on a "compatibility" project.

I only suspect that the Striker implements the outcome of that collaboration. The onboard sound features are named "SupremeFX," which could be a variation of "X-Fi Extreme" chosen to avoid patent-name infringement. I'm only guessing, and offer an observation that I once bought a "Sears" TV that was a rebadged Sanyo.

It may be that an Audigy 2 ZS might work "better," but the source of the problem seems to be less that of compatibility and more a matter of the PCI-E noise.

Several reviewers and forum-posters think the onboard sound option is "just great." Other audiophiles may not HEAR the difference, but insist on buying the Creative Labs X-Fi because "it's the best."

The onboard PWR-ON, RESET and CLR-CMOS switches are a nice enhancement for enthusiast system builders. The RED-GREEN-RED LED lights on those switches provide "bling." The onboard blue LEDs provide "bling-bling." The I/O plate has a built-in LCD back-lit display that reports system-post messages for diagnostic purposes -- essentially adding a PC diagnostic expansion card to the motherboard features.

Personally, I don't regret spending the money, even if it has meant a departure from my practice of finding good, low-to-mid-range moderately priced motherboards to tweak. I don't mind compromising my Creative-Labs inclinations for the onboard sound. I don't see any "instability" in the board. The over-clocking limit seems to relate to heat and the processor. Remarks about "limits" are the same one finds with regard to eVGA and other 680i board-makers.

I'd say, for a 680i board, given the fact that ASUS has been "in the motherboard business" for a longer time than nVidia, this may be the best board available for that chipset.

Oh, yeah. . . One more observation. I'm not running two 8800 GTS in SLI yet -- just a single BFG OC model. But my 3DMark06 v.1.0.2 score is . . . . :

10,765 -- using the 9-multiplier, 360 Mhz external frequency settings. Is that "good enough?" You tell me.



 

BonzaiDuck

Lifer
Jun 30, 2004
16,303
1,872
126
Editorial correction: Eva2000 set the VCORE initially to 1.525V . . . [I erroneously said 2.525 . . . ]

. . . and I just corrected my error in the original post . . .
 

BonzaiDuck

Lifer
Jun 30, 2004
16,303
1,872
126
Footnote #1 [for anyone interested]:

Crucial Ballistix DDR2-1000 modules, under the 360 Mhz external frequency, @ DDR2-720, CPU 3.24 Ghz, rock-solid stability at these latency settings: 3,4,4,8, 2T. CAS 3!!

8 hours, 50 minutes Orthos, before the wind in So-Cal caused a power-outage and started my UPS boxes to chirping -- stopped the test, shut everything down. I figure if it will do 9 hours, it will probably do 12 or 24.

Well, let my put the kibosh on my enthusiasm and let this thread die, if nobody posts with comments.
 

BonzaiDuck

Lifer
Jun 30, 2004
16,303
1,872
126
Thanks, Secretanchitman!

I'm hyped about this board, although I think I probably got an "average" E6600.

But look at this: I updated the 8800 GTS nVidia drivers for my XP MCE. Maybe I imagine, maybe it's just the Arctic Silver5 getting into the "grooves" and chillin' better, maybe it was bumping up my PCI-E #1 to 109 Mhz [See Adrian Rojak's web-site and the BIOS tuning guide, per PCI-E frequency settings and safe over-clocks.]

Temperatures now seem about 2 to 3C lower under load. Whyizzit? Dunno.

3DMark06 this time: 10,801.
 

BonzaiDuck

Lifer
Jun 30, 2004
16,303
1,872
126
Footnote #2:

The "SOO-Preeme FX" onboard audio. I'm just testing this now with a 10-year-old 2.1 Logitech speaker system. It tends to crack a bit at higher volumes. I noticed this, and took time to recall for myself the speaker-setup's shortcomings, to add hope that I wasn't hearing what others complained about.

I'm sure it's the speakers, because I can turn them down just a tad now, and there is no distortion.

Used to play piano and guitar, and probably have the next best thing to "perfect pitch." The SoundMax-with-audio-riser looks reasonably promising.

Will test on 7.1 speaker setup soon and post again.