DFI LanParty UT ICFX3200

[Gary Key]

This screenshot is of current settings with an E6600 at 7x511, Tuniq 120, CPU at 1.5V, OCZ Flex 2GB kit at 2.48V, ATI X1950CF, Seagate 7200.10 320GB x 2, audio enabled, Promise RAID enabled, in other words the system is loaded.

CPU-Z 7x511

The system just passed Orthos Stable at four hours. I will post varying test results throughout the day. However, benchmarks will be in our first preview article shortly. This board has so many options that it will take a couple of articles to properly review it.

 

[Warren21]

Nice! I should easily be able to hit 3.8GHz + on an E6600 with good cooling and proper vCore...

Gary... can you confirm if this is dual x16 CrossFire or x8+x8? There seems to be a lot of confusion due to DFI's spec sheet. Personally, I thought it was dual x16 (this chip is based off RD580, right?).

 

[xero273]

Originally posted by: Warren21
Nice! I should easily be able to hit 3.8GHz + on an E6600 with good cooling and proper vCore...

Gary... can you confirm if this is dual x16 CrossFire or x8+x8? There seems to be a lot of confusion due to DFI's spec sheet. Personally, I thought it was dual x16 (this chip is based off RD580, right?).

Originally posted by: Gary Key
The RD600 core was developed right after the RD480 and before the RD580. It is a hybird chip in essence. It has more in common with the RD480 than RD580.

I'm just glad the Tuniq will fit on the motherboard.

 

[Warren21]

That's not very reassuring to hear.. how many PCIe lanes does RD600 have then?

And if it's x8 + x8 why is ATI/AMD marketing it as CrossFire Xpress 3200... false advertising?

EDIT: Quite frankly, if the 3D performance is up there with the 680i, RD580 and NF590 and past other x8 mode solutions (975X) I could care less about full x16 bandwidth or x8 in dual GPU mode, but still when/if the day comes that they saturate the PCIe 1.0 interface (which I doubt at the rate they're launching PCIe 2.0) it's a little worrisome.

 

[Gary Key]

Just a quick update -

I am running 3-3-3-6 1T at 866 currently with the E6600 at 7x511. This setting will pass every game benchmark, 3DMark, or PCMark we have run including a loop of 3D06 with Nero Recode running in the background converting a 8GB movie down to 4.7GB.

We still have some online game play to complete but the reason I bring this up is that Orthos and Dual 32M SuperPI will each fail within two minutes of being started. Also, under Orthos and the 3DMarks, my 7x511 (3577MHz) reads as 4579MHz, great for screenshots but not so great for accurate readings.

What is more important then? Application stability or Orthos stability? I vote for application stability as my last Orthos stable (6:53 h/m) setting failed three out of five game benchmarks along with 3D03 of all things before I started over.

Also, the chipset is a hybird between the RD480 and RD580, the graphic slots are at X8 electrically and to be quite honest, there does not seem to be an issue with it at 1920x1200 so far.

 

[virtualrain]

Originally posted by: Gary Key
What is more important then? Application stability or Orthos stability? I vote for application stability as my last Orthos stable (6:53 h/m) setting failed three out of five game benchmarks along with 3D03 of all things before I started over.

I'm sad to hear you say that. I think you need to get to the root of the Orthos & Super PI instability. This board and your results are not going to have any credibility without it. Overclockers will ignore any results that aren't proven to be Orthos/PI stable. If the board isn't Orthos/PI stable at any speed, then there is a serious problem and people will stay away from it. If it stops being Orthos/PI stable at a certain FSB or memory overclock, you need to report what that is. Most will consider that to be the max.

So while I tend to agree that application stability is good, it's not enough to satisfy the Overclocking community.

You have to remember that valid or not, Orthos and Dual Super Pi 32M stability are currently THE tests by which stable overclocks are judged to be valid by. If it's not stable, you can't just gloss over it or vote to change the accepted testing methodology for overclocking... you need to find out why it's not stable and/or back it off until it is stable and accept the results.

Remember that Orthos and SuperPI are just doing math... if the system can't do math error free, it's not stable. There is also a reason that Orthos (previously Prime95) is used for stability testing... From the Wikipedia Article...

On an absolutely stable system, Prime95 would run indefinitely. If an error occurs (the tray icon will become yellow from the default red, indicating that the test has halted), there is a chance that the system is unstable. There is an ongoing debate about terms "stable" and "Prime-stable", as Prime95 often fails before the system becomes unstable or crashes in any other application. This is because Prime 95 is designed to subject the CPU to an incredibly intense workload, and to halt when it encounters even one minor error, whereas most normal applications do not stress the CPU anywhere near as much, and will continue to operate unless they encounter a fatal error.

We count on guys like you to dig into this. There is either something wrong with the board, the BIOS, the settings, or your setup (CPU, RAM, etc.), or you are pushing it too far if it's not Orthos/PI stable. Period.

 

[hardwareking]

if the RD600 is dual 8x pci-e why is it called CFX"3200".Shouldn't it be CFX1600?
And if its dual pci-e 8x nvidia won't let ATI live with it saying their 680 i offers superior performance and stuff.

 

[baronzemo78]

Gary, thanks so much for the hard work testing and updating us so far. I for one have been following this mobo for several months and am very excited about it.
One quick question if you have time.

Originally posted by: Gary Key

Also, the chipset is a hybird between the RD480 and RD580, the graphic slots are at X8 electrically and to be quite honest, there does not seem to be an issue with it at 1920x1200 so far.

I plan to game at 1920x1200 on my monitor at home. I know you can't know for certain since the R600 isn't released yet, but assuming it's similar to the 8800GTX SLI (which you mentioned works with hacked drivers) do you think that the 8x crossfire mode will have any effect on R600 crossfire? I'm guessing not, but I value your opinion.

Thanks.

 

[GuitarDaddy]

Originally posted by: Gary Key

What is more important then? Application stability or Orthos stability? I vote for application stability as my last Orthos stable (6:53 h/m) setting failed three out of five game benchmarks along with 3D03 of all things before I started over.

I vote for both!

I have also seen the situation on AMD NF4 chipset where it's completely stable in Prime and SuperPI but not stable in gaming apps, and I've heard reports of the opposite where games pass and Prime/PI fail.

I think this shows that the two different senarios stress the CPU/Chipset in different ways and fail at different points. I know that Prime/PI heavily stress the FPU performance of the CPU and probably don't stress cache much at all, where gaming probably stresses cache more and FPU not so much (just guessing here).

To me the bottom line is to be considered completely stable it must pass both types of testing IMO. The system is only as good as the weakest link, and I personally will never use a system that can't pass everything I can throw at it.

 

[Rike]

Originally posted by: GuitarDaddy
To me the bottom line is to be considered completely stable it must pass both types of testing IMO. The system is only as good as the weakest link, and I personally will never use a system that can't pass everything I can throw at it.

I'd second that. You run programs, they work. Period. That's what stable is.

 

[Gary Key]

Banzai Run -

9x451 Air , lots and lots of air....

Passed Orthos for about 30 minutes before I needed to shutdown to get back to real work not to mention turning those six 120mm fans off. I am deaf now.

 

[Gary Key]

This is what I have been testing at for the preview article.

OC Stable

 

[Rike]

Originally posted by: Warren21
Nice! I should easily be able to hit 3.8GHz + on an E6600 with good cooling and proper vCore...

Hmmmm . . . as predicted.

Hey Gary, will Anandtech use any water setups to test extreme stuff in the future? I understand why you wouldn't (there are so many more variables with a water setup vs. air), but I thought I'd ask.

In any case, keep up the good work. I'm still looking forward to your pre/review of the board.

 

[Gary Key]

Originally posted by: virtualrain

Originally posted by: Gary Key
What is more important then? Application stability or Orthos stability? I vote for application stability as my last Orthos stable (6:53 h/m) setting failed three out of five game benchmarks along with 3D03 of all things before I started over.

I'm sad to hear you say that. I think you need to get to the root of the Orthos & Super PI instability. This board and your results are not going to have any credibility without it. Overclockers will ignore any results that aren't proven to be Orthos/PI stable. If the board isn't Orthos/PI stable at any speed, then there is a serious problem and people will stay away from it. If it stops being Orthos/PI stable at a certain FSB or memory overclock, you need to report what that is. Most will consider that to be the max.

So while I tend to agree that application stability is good, it's not enough to satisfy the Overclocking community.

You have to remember that valid or not, Orthos and Dual Super Pi 32M stability are currently THE tests by which stable overclocks are judged to be valid by. If it's not stable, you can't just gloss over it or vote to change the accepted testing methodology for overclocking... you need to find out why it's not stable and/or back it off until it is stable and accept the results.

Remember that Orthos and SuperPI are just doing math... if the system can't do math error free, it's not stable. There is also a reason that Orthos (previously Prime95) is used for stability testing... From the Wikipedia Article...

On an absolutely stable system, Prime95 would run indefinitely. If an error occurs (the tray icon will become yellow from the default red, indicating that the test has halted), there is a chance that the system is unstable. There is an ongoing debate about terms "stable" and "Prime-stable", as Prime95 often fails before the system becomes unstable or crashes in any other application. This is because Prime 95 is designed to subject the CPU to an incredibly intense workload, and to halt when it encounters even one minor error, whereas most normal applications do not stress the CPU anywhere near as much, and will continue to operate unless they encounter a fatal error.

We count on guys like you to dig into this. There is either something wrong with the board, the BIOS, the settings, or your setup (CPU, RAM, etc.), or you are pushing it too far if it's not Orthos/PI stable. Period.

1. I am not glossing over the importance of Orthos although SuperPI is not that stressful on a system. The issue I am raising and raised a few years ago with Prime is that the program needs an update in my opinion based upon the results we have seen seen since the launch of Core Duo and now Core 2 Duo. We have new CPUs and chipsets now that operate differently than what was available when the programs were last updated.

2. Prime and MemTest would not run correctly on the Athlons when they were released and the programs were updated several times for P4 and A64. They are in need of another update at this time. If we ignore the fact that we are relying on software that has to be modified to account for new hardware designs then we will be short changing ourselves.

3. See for yourself the number of program changes that occured with the Athlon and P4 processors - Prime95 Program Updates - and then tell me if you have 100% confidence in a program that has not been updated for the current series of processors. This situation gets even worse with Conroe+ on Bearlake.

4. I have already shown that I can get Orthos stable with this system but it will fail seven of twenty six benchmarks. I do not know about you but I would not call that system stable.

5. I have pushed the system to no end, but I am sorry, Orthos stable does not cut it for me right now although our final settings will be Orthos and application stable with the notation that it is time to update the programs we use to determine stability again. It could be the SB600 driver set is causing an issue, it could a number of things, but I can only report on what I experience and so far that leads me to believe there are more issues with Orthos/Prime95 than there are with my settings.

6. Also, what does Orthos tell us about the stability of your video, network, or audio component. I would love to see a program that covers all areas. In the meantime, I play games, watch videos, create music, and do office work on my PC. Not one time have I sat down and thought it would be nice to play Orthos or have Prime95 do my taxes.

 

[vladman]

Gary, sorry if I sound like a noob (I'm really not, but never had a C2D E6600). Are you able to adjust the multiplier with this mobo? I though these CPUs were locked???
I plan on getting an E6600 with this mobo.
Thanks - this combo looks like a real winner.

 

[xero273]

The multi can be adjusted to be lower.

 

[nealh]

I have to agree with Gary here..app stablity is key

While I like a system to be P95 stable I have seen too many issues with how long ..then someone posts 24hr P95 stable but fails games immediately(I am not talking even overclocked)

Too many variables in a computer in some cases

Hardware and software can lead to crashes, with that said Prime is "maybe" the best we have for now

If a computer does not crash with the tasks I do..I am happy...I still do all my ovcerclocking testing with Prime and I like no less than 12-15hrs without a crash....

I am using a 1.5 yr old winxp install and I have the new hardware wizard popping up every boot..ugh..so software issues can be a bitch to work..I hope the authors of Prime see if an update will help

 

[virtualrain]

I think we are all on the same page... more or less... with respect to Orthos/Prime95 stability.

Yes, it may need some updating, but when you think about it, all it is doing is mathematical calculations and is one of the few programs that actually checks to see if the results are correct... Do you think your apps and benches give a crap if the results are correct... no!

Furthermore, the overclocking community needs SOME benchmark or stress test to say "this overclock is stable". Prime95 and Orthos ARE that test and you aren't just going to dismiss it overnight. Sure it may need an update, but on the other hand, maybe it doesn't. Maybe the DFI board or your proc or your memory actually has issues!

While every seasoned overclocker knows that you also test with a variety of other apps and benches, the simple fact is that if it doesn't pass an 8 hr test of Orthos/Prime, it's not even worth doing other tests.

The reason it has been the test of choice is because it is a) very stressful and b) fails on any calculation error. No other benchmark/test I'm aware of does this. If you have another one, let's use it instead.

Any seasoned overclocker knows that you can overclock to a point where Windows and apps will run fine but Orthos will fail. That's exactly why we all rely on the "Orthos" test because it doesn't mask anything.

Having said all that, if it needs an update, great... let's do that, but how do we know it's Orthos fault and not the motherboards, CPU's or RAM's fault it can't do math accurately?

 

[virtualrain]

You will notice I'm pretty passionate about this... I'm not sure why... If everyong thinks their overclock is stable after a couple of passes of 3DMark06, I guess that's up to you. However, I feel some desire to have a litmus test for overclocking stability, otherwise, the whole hobby is thrown into chaos... without a firm test like Orthos/Prime, how can you tell a good overclock from a bad one? How can you really know good RAM timings from bad? It's pure chaos and absolutely rediculous what will happen without a firm bench/test that everyone can believe is the statement of a stable overclock.

I find that some people's enthusiasm for this over-hyped DFI board is causing them to throw best practices out the window.

 

[Imaginer]

Glad to hear some performance tidbits about this board.... However I am still crushed that the slots are only 8x in crossfire mode.... The fact that I thought it was a full 16x for both slots was something I looked forward to in this board. Now I am not sure if I can justify the price point. Maybe if they threw in two more SATA ports... or something. Also I am certain this board does not support Core 2 Quads (or am I wrong on this one).

If not, then its even a harder thing to bite... I am torn whether or not to get this board.

 

[nealh]

Originally posted by: virtualrain
You will notice I'm pretty passionate about this... I'm not sure why... If everyong thinks their overclock is stable after a couple of passes of 3DMark06, I guess that's up to you. However, I feel some desire to have a litmus test for overclocking stability, otherwise, the whole hobby is thrown into chaos... without a firm test like Orthos/Prime, how can you tell a good overclock from a bad one? How can you really know good RAM timings from bad? It's pure chaos and absolutely rediculous what will happen without a firm bench/test that everyone can believe is the statement of a stable overclock.

I find that some people's enthusiasm for this over-hyped DFI board is causing them to throw best practices out the window.

I do not disagree or dismiss you points and issues..they are all valid

But Orthos/P95 do not guarantee stabilty of your machine for everyday use, gaming etc(I am talking even stock..not overclocked..how often do we see a Dell person post that they are failing P95 more than once)

There maybe an issue with the chipset or whatever....if it is hardware, ie the chipset..this is a issue and not sure how someone fixes it.....it maybe just too many options that are somewhat aggressive that causes an issue

I used to run a P42.4b@3.0 ghz for like 3yrs....it was not P95 stable(did not really care 4yrs ago)..it did everything without issue for my needs and I was happy

After that build..everyone has been P95 stable for at least 12hrs(which meets my needs...I just stop the P95 run...they have not failed)

 

[Warren21]

One more day! I just hope availability is decent. Can you give us any word Gary (since you were working with DFI)

 

[xero273]

Here's what Tony @ XS has been saying about availability of the board.

http://www.xtremesystems.org/forums/showpost.php?p=1893224&postcount=443

Expect high prices, but I'll still try to grab one.

 

[Skott]

Even at high prices availbility is going to be scarce till January if the rumos of only 100 mobos slated for U.S. sales holds true. One upside to this is by the time the mobos does become more easily obtainable BIOS updates should be availble and in place and the early users should have a overclocking path blazed for the rest of us. This mobo is going to have a steep learning curve methinks.

 

[Praxis1452]

Originally posted by: virtualrain
I think we are all on the same page... more or less... with respect to Orthos/Prime95 stability.

Yes, it may need some updating, but when you think about it, all it is doing is mathematical calculations and is one of the few programs that actually checks to see if the results are correct... Do you think your apps and benches give a crap if the results are correct... no!

Furthermore, the overclocking community needs SOME benchmark or stress test to say "this overclock is stable". Prime95 and Orthos ARE that test and you aren't just going to dismiss it overnight. Sure it may need an update, but on the other hand, maybe it doesn't. Maybe the DFI board or your proc or your memory actually has issues!

While every seasoned overclocker knows that you also test with a variety of other apps and benches, the simple fact is that if it doesn't pass an 8 hr test of Orthos/Prime, it's not even worth doing other tests.

The reason it has been the test of choice is because it is a) very stressful and b) fails on any calculation error. No other benchmark/test I'm aware of does this. If you have another one, let's use it instead.

Any seasoned overclocker knows that you can overclock to a point where Windows and apps will run fine but Orthos will fail. That's exactly why we all rely on the "Orthos" test because it doesn't mask anything.

Having said all that, if it needs an update, great... let's do that, but how do we know it's Orthos fault and not the motherboards, CPU's or RAM's fault it can't do math accurately?

When your overclocking you are overclocking the whole point is to get a better benchmark of faster fps or something to that extent in apps.

Nobody benches for Orthos because it doesnt' really give a score. Nobody should care if your overclock is Orthose stable because if you can bench at 4.5ghz instead of 4.3ghz orthos stable then it's much better. Orthos stability was only used because it was supposed to gurantee app stability. If your apps are heavily stressing as 3dMark06 is stressful or BF2(poorly coded) and they still work fine then it's stable. No if and's or buts. Orthos stable, maybe not but unless you are benching for orthose hours stable then it is absolutely useless.

Many overclockers do 1M pi runs and no they don't check orthose for 8 hours stability.

In this case Orthos is failing while all others are passing. When orthos passed for 6:53 mins I believe 7 apps failed. In reality orthos then means nothing. It's not even a benchmark for stabilty.