ASUS Striker Extreme 1301 BIOS 7/11 Update

[hyde211]

Beta Version： 1303 2007/07/26 update

OS： DOS / All OS

Description： Striker Extreme BIOS version 1303
1. Enhance CPU overclock capability of FSB1333 processors
2. Add new CPU uCode to support new processors
3. Enhance memory compatibility

File Size： 668.32 (KBytes)

Global

[slim142]

thanks for the changelog! now I see reasons to update!

[slim142]

I hate my E6600 needs 1.42500 to do 350mhz FSB

I know is not the board since its a bad LA628A

[Faiakes]

I just received my E6750 (L719B030), testing now with 1303 BIOS.

UPDATE 1:
I've already managed to achieve 450 FSB with memory Linked:Sync Mode

vCore: 1.5 (it did that on Auto too)
HT: 1.4 (from Auto)
NB: 1.4
SB: 1.55
VTT: 1.45

(all figures in BIOS)

I am testing for stability now and will report results later today.

UPDATE 2:
I have managed a stable 480 FSB and and boot into windows with a 500 FSB but I need to tweak it to make it stable.

The E6750 reached 3.6GHz with 8x450, 1.525 in BIOS (1.5 in Windows) Idle temp at 24 C. Will further tweak it.

[markost]

Unfortunately neither 1301 nor 1303 fix the problem of memory performance dropping after 424 FSB. Seems that the striker is brick walled performance wise at that point. Everest read and latency benches drop significantly after that point. Also is it just my quirky board or does 1303 react adversely with Razer Copperhead Mice. I can't get passed the first BIOS screen with my mouse plugged in. Flash back to either 1102 or 1301 and no problems. 1102 is still the quickest BIOS at least when used with OCZ VX2 or Reaper dimms, E6700 @424 x 10. Dropping in an E6850 over the next couple of days so that may show if the 424 FSB performance brick wall is fixed with G0 chips. 9X multi may be what holds it back . Unfortunately you need phase to get an E6700 over 4.0 Ghz using 10x multi and then the board runs out of steam at 424 FSB. Phase on a Striker/E6850 would be a waste of time as the 9X multi is too low. Keeping memory at the same speed, 500 FSB is decidedly slower than 424 FSB (memory wise). What ASUS really need to do is write a BIOS that doesn't choke memory performance after 424. Will this ever happen ???? If it wasn't for SLi, I would have been out of this board ages ago.

[Faiakes]

Really? 424 FSB same as 500?
So the only benefit is the higher CPU clock?

[markost]

Really? 424 FSB same as 500?
So the only benefit is the higher CPU clock?

Yeah mate. 500 is slower than 424. Did the same on another phase machine using a X6800 with the same results. He actually dropped back from 500 x 9 to 424 x 10 (-260 Mhz) and prefers it there. We all know that the memory interface is what limits speed so why would you build a reference board that slows down at the top end. Sure 3D and synthetic CPU benches will always be better with high FSB, but the idea if this whole exercise is to get the most physical speed possible. You can't get that when a board actually degrades performance after a certain point. I've built heaps of high end machines for clients and often the fastest performance wise (not bench wise) are the ones which have the memory interface tweaked to the eyeballs.

A Porsche GT3 RS has nowhere near the HP of a Bentley Continental but gets around a race track a damn site quicker.

[Faiakes]

How did you measure that? What kind of software/tests? You got me intrigued now.

Has anyone had the same experience?

[Leeghoofd]

yep, that's why I find it so amusing that so many peeps are pushing their FSB and ram to extreme limits while it doesn't neccesarily give that better performance (due to looser strap, timings)... for me I'm running always in safe mode in comparison to other members here

I run my E6600's at 3.6Ghz for me 9 x 400, DDR 800 3-3-3-10 1T is faster than 8 x 450 , DDR 1066 5-5-5-15 2T

you can measure ram bandwith with memtest, everest home, Sisoft Sandra and co, SuperPI, 3dmarks etc....

I did some quick tests for you keeping ram at the same speed (DDR800 with 4-4-4-5), I get more gain running with 1T then running high ram clocks with 2T (never been over 1120mhz)

E6600@3600Mhz
9 x 400 , 8 x 450 ram DR800 4-4-4-5 2T, if I set ram at 1T at 450FSB I get stability issues 3rd 3600Mhz DDR800 4-4-4-5 1T


Everest read 10238 --- 10116 --- 10660
Everest Write 7306 --- 8214 --- 7305
Everest Copy 6783 --- 6711 --- 7296
Latency 52 --- 52.9 --- 48.7

Aquamark 185983 ---186063 --- 192737

SuperPI 1m 14sec 306 --- 14sec 313 --- 14sec 250


It's hard though to compare some numbers with other's as some peeps use an older or newer version of the software... but you can compare for yourself if a certain value gives you that extra ooomph....


I think the Asus board handles tighter timings then the EVGA and BFG reference ones, hence why we are blocked with lower FSB OC in general and the board is also harder to setup then the mentioned ones.... and I can confirm if you didn't need SLI don't buy a 680i it's a waste of money, P35 feels so much smoother and is so much easier to setup...

And there is a problem with USB copperhead mice as some boards fail to boot when plugged in... apparently older biosses don't have this issue... (reminds me of the old Abit AN8 SLI days lol)

[markost]

How did you measure that? What kind of software/tests? You got me intrigued now.

Has anyone had the same experience?

Hey Faiakes. Leeghoofd sums it up pretty well in the post above. Use all the well known benching apps and they'll all tell you the same story about the Striker. In terms of speed, your wasting your time going higher than 424 FSB. If you don't have SLi, don't waste your time with the Striker. But as far as SLi boards go, it's about as good as you can get. I haven't tried the DFI 680i LT and from what bigtoe is doing with it http://www.thetechrepository.com/showthread.php?t=157
it may be the quickest but not with quads.

[pmp]

Is this what pmp called BIOS 1401?

No m8. It's not. Currently next major upgrade is still under qtc, just be patient
More details by beginning of next week, when i finish my vacation time.

[markost]

No m8. It's not. Currently next major upgrade is still under qtc, just be patient
More details by beginning of next week, when i finish my vacation time.

Good stuff PMP. just when you think not much is happening, a carrot gets dangled to keep you keen. Are you hinting that the huge performance drop after 425 FSB will be fixed up ?????

[Faiakes]

I see.

Well, I definitely want SLI which is I why I forked out for an 680i board.

I have currently reched a 8x475=3800 (3DMark06 11993) with my E6750 and I'll run Everest and Sandra to confirm the above mentioned memory performance limit.

By the way the DFI board seems to have a lot of issues. It is nowhere near the Striker (nice chipset cooling scheme though)

[Leeghoofd]

Well those settings are nice Faiakes but I wouldn't recommend running 475 FSB 24/7... I back off to 333FSB for day to day use and only when I need it (benching, showing off) I load overclock setting 2

[Faiakes]

Of course not. That is just for playing STALKER at max settings (damn game!)
Although load temp is 48 C, which isn't bad. If only I could learn how to spead that paste properly...

I am back to profile 1 with stock settings for cool, quiet and economical operation. Now I am interested in upgrading my memory to something as good as yours though. Care to provide a link and some info?

[Faiakes]

OCZ PC8500 800mhz 3-3-3-9 1T 2.25 volts - - GSKILL PC8000 1000mhz 4-4-4-5 2T 2.25 volts

About your memory:

I have found the GSkill online but my 7200 can already do 4-4-4-12 up to 500MHz. Would there be any performance benefit from upgrading to the GSkill one?

Your PC(2)8500 is very impressive. How likely would it be that I could match that if I bought a pair of those too?

[Leeghoofd]

I'll do some tests with the current Gskill PC6400PK versions I test (4-4-4-12 and 4-4-3-5) to do some tests at stock ram settings... as these rams ahve almos tno headroom to overclock but great for day to day use.......

If your OCZ 7200 can reach OCZ 8500 plat timings mate 1000Mhz at 4-4-4-12 I would keep them but at how many volts is that ? 2.3 ? I would surely recommend added air cooling... Also noted that the expensive ram sometimes isn't really worth the hassle , I got the Gskills, as they were only 30 euro's more expensive than the OCZ 8500SLI I already owned... but performance difference on general applications is barely noticeable.... great for SuperPi and 3d benching though....

[Faiakes]

At officially approved 2.2v

Surely 3-3-3-9 1T must be a good performance leap over 4-4-4-12 2T ?

I am only interested in game perfromance.

[1Day]

Faiakes I seriously doubt that you will notice any real-world (game) improvements between the two different ram timings. By that I mean visual difference of any noticeable kind. However in benchmarking there will be a measurable difference, in so much that having a extra process running in the background (LAN connection for example) will demonstrate a measurable difference too.

[Leeghoofd]

Well you would be surprised how little difference fast ram makes, running 1T still has the biggest effect on 3dbenchies and co but little effect on gaming...

I don't own many games to test only play Quake 3/4 and GTR2 online... but as these are already older games no idea how to give ya decent output...

If I was you I would try to get all in sync meaning eg 450FSB and 900mhz ram... if you want to drop in CPU performance but gain a bit more in ram bandwith and co go to 400 region and run the ram in sync with the lowest timings you can and 1T to get the fastest ram performance (according synthetic benchies) I think you can go DDR 800 4-3-3-10 1T surely with the 7200 with 2.2 - 2.3 volts... My 2nd rig is off to a lan party with my nephew ( with the OCZ in it ) so I cannot test them till after the weekend...

[Faiakes]

OK thanks for the advice. I suspected as much. I had read a couple of articles from Xbit Labs and Tom's Hardware suggesting the same thing.

On a different note, I tried to recreate the RAM performance issue so I have been running some tests.

I can't recreate the issue (as far as I know):

1) RAM 424 (x8), Linked, Sync Mode, 4-4-4-12-13 2T, CPU = 3392
Everest Ult. 4 = 10653 7737 7034
Sandra 6 = Combined 22206
RMMA 3.72 = 3426

2) RAM 485 (x7), Linked, Sync Mode, 4-4-4-12-13 2T, CPU = 3395
Everest Ult. 4 = 11335 8833 7478
Sandra 6 = Combined 23140
RMMA 3.72 = 3831

That is the closest match I could find between CPU speeds.

It would seem there is a linear increase in RAM performance. Am I missing something?

[Leeghoofd]

Well you have to compare ram at the same speed so you need to force eg DDR800, now you are running ram at 860 or so and once close too 1000.. can't you run ram at the first setting 1T ?

[Ambro]

Well those settings are nice Faiakes but I wouldn't recommend running 475 FSB 24/7... I back off to 333FSB for day to day use and only when I need it (benching, showing off) I load overclock setting 2

get some decent skills + cooling

my stock is 500 FSB - with low volts & 1303

I'm waiting for the next bios to fix some "holes" (won't run between 1680 - 1720, etc.)

new found faith with this bios release though

[Faiakes]

Oooops, you are right. i was measuring purely RAM when I ought to have set the RAM at a given speed and only run the FSB.

Will get right to it.

[Faiakes]

OK, redid the test and these are the results I got:

1. FSB 424, 16:15, Unlinked, (397 RAM), CPU = 2967
- Everest = 9574 7723 6375
- Sandra = 19871
- RMMA = 3185

2. FSB 475, 6:5, Unlinked, (395 RAM), CPU = 2850
- Everest = 8247 7574 6224
- Sandra = 19300
- RMMA = 3173

So, yes. Despite the extra 51MHz of the FSB there is actually a decrease in performance, which is disturbing as it negates the reason for getting this board.

Unless you wish to run the memory at 1:1, in which case:
3. FSB 475, 1:1, Linked, Sync Mode, CPU = 3800
- Everest = 11929 8761 7406
- Sandra = 24784
- RMMA = 3797

UPDATE:
I compared same FSB but different memory speeds but got somewhat conflicting results:

1. FSB 424, 1:1, CPU = 2543
- Everest = 10046 6767 6731
- Sandra = 18477
- RMMA = 3384

2. FSB 424, RAM 475 (actual 471), 9:10, CPU = 2543
- Everest = 9930 6763 6939 ( Read Write Copy)
- Sandra = 19202
- RMMA = 3683

So, there seems to be a better result on Copy, the Sandra and RMMA but a decrease in Read and Write performance.