Do the FX chips really have a bug in them?

[sequoia464]

Then why skip the first gen? Seems like that'd be more fun as a hobby if not as useful in daily usage :biggrin:

I probably will pick up an AM3+ board + FX-8xxx if there are decent sales during black friday... because why not? I really just want to see how high an overclock I can get w/out having to worry about spending a ton of money if I blow it up... ATM it is priced too high for that, but once the supply issues are worked out, I expect the prices will drop.

These things are different than the Phenoms and Thubans to overclock.

I have a 6100 that I am struggling with. Memory seems to be a very important factor for me, have some Crucial as well as Kingston that are both on ASUS's approved list but memory settings seem to be, at least for me, much more important than with a Phenom.

What I have done to get stable is underclock the memory - raise the FSB but leave the multiplier and turbo core alone. This is giving me 3630MHz normal and 4290MHz when the turbo core kicks in. This means that my 1600 memory is only running at 1470 MHz - I had to pull two sticks as well so I'm down to 8GB.

I know most will disable the turbo core function, but this setup works fine for my usage.

I have never struggled as much with an overclock, I'm sure my problems are operator error, just don't have things figured out yet - I have seen some decent overclocks with the 6100. I'm not completely comfortable with my Sabertooth's bios yet either.

I got the 6100 for $160 or I probably wouldn't have bought it.

 

[Kalessian]

if he got him a thuban x4 then he's still good to go

What's the unlock chance/rate on those?

 

[Chiropteran]

What's the unlock chance/rate on those?

They either unlock or they don't, so it is 50%

 

[nanaki333]

yep, luck of the draw just like the old phenom x2 and athlon/phenom x3

 

[cebalrai]

I highly doubt there's a bug. A friend at work has been playing Shogun 2 on an FX-6100 and hasn't mentioned any problems.

 

[exar333]

They either unlock or they don't, so it is 50%

No

Those are the two results of the unlock. I have not heard the actual percentage to unlock was 50%. Link?

 

[bryanW1995]

The respin did more than just fix the bug. It was also for public image to fix it. It increased yields and especially clockspeed. The B3 respin got them up to 9950 models unlike B2 and 3ghz+ oc's. So from My pov, they were able to do more than just scramble to fix a bug only respin. B3 was coming anyways. It was in their best interest to fix it in the new revision. If you remember it was found out in-house not because 10000 customers were complaining. And of course they are going to fix errata that they deem as needing fixed. Internal testing have showed that its in their best interest to fix it.

Again, my opinion, I Never claimed I have inside info or anything. If you don't agree, thats fine. I am just going by what the new respin brought aside from the TLB bug fix.

And since neither of us truly knows the entire story, its only opinion based on some evidence anyways.

So in the end, My opinion is: It was a majorly overblown bug. Thats all. Again just opinion. I'm also pretty much done with this thread. PM would be ok.

Looks like he didn't get a buggy BD afterall.

I see the TLB bug as similar to intel's issues earlier this year with the SB launch. The only difference is that intel could afford to stumble b/c AMD wasn't poised to take advantage of it, while the TLB bug was at a critical juncture as it was AMD's (very late) response to Conroe. Intel's main competition for SB this spring was Nehalem.

These things are different than the Phenoms and Thubans to overclock.

I have a 6100 that I am struggling with. Memory seems to be a very important factor for me, have some Crucial as well as Kingston that are both on ASUS's approved list but memory settings seem to be, at least for me, much more important than with a Phenom.

What I have done to get stable is underclock the memory - raise the FSB but leave the multiplier and turbo core alone. This is giving me 3630MHz normal and 4290MHz when the turbo core kicks in. This means that my 1600 memory is only running at 1470 MHz - I had to pull two sticks as well so I'm down to 8GB.

I know most will disable the turbo core function, but this setup works fine for my usage.

I have never struggled as much with an overclock, I'm sure my problems are operator error, just don't have things figured out yet - I have seen some decent overclocks with the 6100. I'm not completely comfortable with my Sabertooth's bios yet either.

I got the 6100 for $160 or I probably wouldn't have bought it.

Average improvement on 600 mhz turbo is ~ 150 mhz, just OC it to 3830+ with turbo off. Interesting comment about the memory, though not surprising unfortunately.

 

[sequoia464]

Average improvement on 600 mhz turbo is ~ 150 mhz, just OC it to 3830+ with turbo off. Interesting comment about the memory, though not surprising unfortunately.

I'll try that - thanks. Also looking at some G.Skill lower latency, lower voltage memory. Kind of pricey (in comparison), but it might help.
http://www.newegg.com/Product/Produc...82E16820231404

 

[waffleironhead]

They either unlock or they don't, so it is 50%

Is that like the odds of winning the lottery? You either win or you dont? 50%

 

[podspi]

I have never struggled as much with an overclock, I'm sure my problems are operator error, just don't have things figured out yet - I have seen some decent overclocks with the 6100. I'm not completely comfortable with my Sabertooth's bios yet either.

I got the 6100 for $160 or I probably wouldn't have bought it.

This is what I mean, though. If you're buying the computer mainly to get work done, this is annoying. If you're buying it as a hobbyist, if its too easy what's the point? That's why so many enthusiasts have picked up Llanos. They really aren't fast per-se, but they're fun to tweak!

I was hoping for some great intro-pricing on the FX-series (like there was with the X6s which I foolishly held out on), but alas it wasn't to be.

I've been meaning to ask the first person who claimed to have one of the 6 or 4 core FXs. Have you tried core/module unlocking yet? I know AMD said it wasn't possible, but they've said that before...

Is that like the odds of winning the lottery? You either win or you dont? 50%

OT: Uh-oh, is this going to devolve into another math thread? :whiste:

 

[Digital Savior]

Just tell him hyperthreading gives hims more cores!

 

[exar333]

Is that like the odds of winning the lottery? You either win or you dont? 50%

By that logic, my CPU either OCs to 10ghz or it doesn't. 50% chance to get 10ghz!!!! SWEET!!!

 

[Arachnotronic]

They either unlock or they don't, so it is 50%

Unlock rate = # unlocks/#attempted unlocks. Learn basic probability

 

[soccerballtux]

Unless you work for AMD's RMA dept, what you don't know is how widespread the problem was for AMD's customers. You have your opinion of it and that is all.

Meanwhile, AMD had their data along with their customer reports. Based on their facts, not opinion and speculation, they decided spending millions on a respin to address the bug versus spending those millions to improve performance was the best path forward.

Surely you can see how absurd it is that you would hold yourself out as knowing more about the depth and magnitude of the issue than AMD?

Evidently, despite what you think to be true or false about the TLB situation, AMD's decision (made with the benefit of actual data on the impact to customers) suggests they had their reasons for doing what they did.

You might never know their reasons, just as you might never know the true extent of the customer impact caused by the TLB bug, but you do know that a company felt it was a necessary step and that alone is very telling.

that is one way of looking at it but given amd's general track record (pursuit of monolithic quad core, not firing Ruinz earlier, etc)....just sayin.

That said, since it affected server parts, I agree.

 

[PreferLinux]

By that logic, my CPU either OCs to 10ghz or it doesn't. 50% chance to get 10ghz!!!! SWEET!!!

Don't you mean 100 GHz??? Or even 1000?

 

[Chiropteran]

It was a joke. I don't think anyone really knows the unlock rate yet.

 

[FalconHorse]

http://www.xtremesystems.org/forums...lve-CEG-games-causing-BSOD-on-Bulldozer-issue

Give the man his 8 cores!

 

[jordanecmusic]

So, I need to do a build for a client. He's heard all about these 8 core FX processors and he wants one. He doesn't care about benchmark data or anything like that. He just wants to be the first on his block with an "8 core CPU".

That's fine. I'll build it, no problem. However, I have to know if these FX chips are actually as buggy as I've been reading. If this is the case, then I will refuse to do the build (assuming my client does not want to give up "8 COREZ")

In Other News: Man Believes The Internet! Story at 10!

Oh Baby Harder!

 

[VirtualLarry]

Hmm. I'm running PrimeGrid through the BOINC client, and I just recieved this message:

PrimeGrid: LLR v3.8.6 released
After experiencing some erratic behavior from several new hosts, specifically the AMD FX(tm)-8120 Eight-Core Processor, we have updated LLR from v3.8.4 to v3.8.6. This latest version uses GWNUM v26.6, and it appears to have solved the problem.

Ok, so this proves that the AMD FX CPUs DO have issues with existing code compatibility, and require workarounds. Seems buggy to me.