Phenom II x3 710 run with 4 core!

[geokilla]

If I had the money, I would definitely get myself a Phenom II X3 setup and test this out. I could care less if my Windows gets corrupted or not because I'd have everything backed up onto my other computer. Then I'd test the Phenom rig to make sure that it's all stable and everything. It's not like I have anything important in my computer besides some music and homework files.

So besides Mark, is anyone willing to try it out here?

 

[Markfw]

Originally posted by: SunnyD

Originally posted by: Denithor
Couldn't be as bad as Probabilistic CPUs.

I can see those corrupting your data in a matter of minutes, not weeks/months.

Besides, if you unlock that fourth core and run linpack/occt/etc without errors - wouldn't that prove it's fully functional & stable?

No, because each of those only test subsets of instructions/ALU's/registers/cache, etc. For example, I was testing my 940BE... was prime stable. Started up F@H and my system instantly rebooted.

Overclocked to much, or not enough PSU. F@H is a great stability checker

And I am not the guinia pig, Flipped Gazelle said he has one coming and will try it out.

 

[SunnyD]

Originally posted by: Markfw900

Originally posted by: SunnyD

Originally posted by: Denithor
Couldn't be as bad as Probabilistic CPUs.

I can see those corrupting your data in a matter of minutes, not weeks/months.

Besides, if you unlock that fourth core and run linpack/occt/etc without errors - wouldn't that prove it's fully functional & stable?

No, because each of those only test subsets of instructions/ALU's/registers/cache, etc. For example, I was testing my 940BE... was prime stable. Started up F@H and my system instantly rebooted.

Overclocked to much, or not enough PSU. F@H is a great stability checker

And I am not the guinia pig, Flipped Gazelle said he has one coming and will try it out.

That was at 3.5GHz at stock voltage. PSU is a Corsair TX850W. It could also be an issue with F@H too, since I haven't a clue what I was doing with it.

 

[Markfw]

Originally posted by: SunnyD

Originally posted by: Markfw900

Originally posted by: SunnyD

Originally posted by: Denithor
Couldn't be as bad as Probabilistic CPUs.

I can see those corrupting your data in a matter of minutes, not weeks/months.

Besides, if you unlock that fourth core and run linpack/occt/etc without errors - wouldn't that prove it's fully functional & stable?

No, because each of those only test subsets of instructions/ALU's/registers/cache, etc. For example, I was testing my 940BE... was prime stable. Started up F@H and my system instantly rebooted.

Overclocked to much, or not enough PSU. F@H is a great stability checker

And I am not the guinia pig, Flipped Gazelle said he has one coming and will try it out.

That was at 3.5GHz at stock voltage. PSU is a Corsair TX850W. It could also be an issue with F@H too, since I haven't a clue what I was doing with it.

Stock voltage is not enough....I can install windows@3.8, but its not stable F@H except@3.6@1.4 vcore. If it was software, it just would EUE.

 

[allies]

SunnyD - How long was your CPU prime stable?

Flipped Gazelle - impatiently awaiting for your results...

 

[magreen]

Originally posted by: Zstream

Originally posted by: MarcVenice
Idontcare, what makes you so sure the X3's fourth core is dysfunctional? Do you work for AMD? Did they tell you in a official statement the fourth core was gone bad, so they made it an x3, instead of an x4?

I'm not sure, maybe you're right, but this can't be compared to overclocking. There's thousands of ppl who overclock, and I've never heard any horror story's about corrupted files. That would mean that a orthos prime test is a bad way of measuring the core's stability or instability. Coz orthos prima, intels burn test etc seem to be working just fine in determining your overclocks stability.

He doesn't know. Just him making smoke as usual.

GO. AWAY. TROLL.

 

[magreen]

Originally posted by: MarcVenice
Idontcare, what makes you so sure the X3's fourth core is dysfunctional? Do you work for AMD? Did they tell you in a official statement the fourth core was gone bad, so they made it an x3, instead of an x4?

Nothing makes him so sure. idc didn't say he was sure. He clearly said that some non-zero percentage of x3s have a damaged 4th core, just as some non-zero percentage of x3s have a perfectly functional 4th core. I don't see why you take the tone you do in your post. Here's what he said:

Originally posted by: Idontcare

Originally posted by: error8
How do you know it's damaged? It might be a perfectly functional core. I don't think that AMD makes X3s just from defective quads. They probably also use healthy quads with one core deactivated.

That's not the point, the point is a non-zero percentage of X3's are in fact harvested X4's with dysfunctional cores. Not all damaged cores are going to give you an obvious system crash or instability, pentium FDIV bug for example.

There will be those incipient errors that slowly accumulate and creep into all your files (defrag touches them all eventually) and you start pulling your hair out over the "file damaged, cannot be read, CRC invalid, etc" alerts.

So simply enabling a core and claiming "w00t! she's p95 stable" does not mean your spreadsheets aren't about to become fubar, or your files of family photos getting corrupted as they are written to the hdrive, etc.

The caveats are no different than for your regular overclocking enthusiasts, but enabling what can very well be a walking wounded core just seems like extraordinary level of risk taking.

 

[taltamir]

maybe the appliance manufacturers could "corner the market" by prescratching their brand new appliances... ooh, and honda can hire people to beat new cars with bats to corner the low end "damaged car" market.

Anyways, if it passes linpack then it is NOT necessarily working correctly with no errors... that is just one test out of a barrage of needed ones. I don't have the full list because I am not a CPU engineer.

 

[VirtualLarry]

Originally posted by: taltamir
maybe the appliance manufacturers could "corner the market" by prescratching their brand new appliances... ooh, and honda can hire people to beat new cars with bats to corner the low end "damaged car" market.

Anyways, if it passes linpack then it is NOT necessarily working correctly with no errors... that is just one test out of a barrage of needed ones. I don't have the full list because I am not a CPU engineer.

There is a market for pre-worn jeans...

 

[dmens]

Anyways, if it passes linpack then it is NOT necessarily working correctly with no errors... that is just one test out of a barrage of needed ones. I don't have the full list because I am not a CPU engineer.

reminds me of a funny story, one of the most unstable programs on prescott was the windows xp screensaver where the tetragon bounces around. it was so bad the thing got incorporated into production testing.

 

[taltamir]

Originally posted by: VirtualLarry

Originally posted by: taltamir
maybe the appliance manufacturers could "corner the market" by prescratching their brand new appliances... ooh, and honda can hire people to beat new cars with bats to corner the low end "damaged car" market.

Anyways, if it passes linpack then it is NOT necessarily working correctly with no errors... that is just one test out of a barrage of needed ones. I don't have the full list because I am not a CPU engineer.

There is a market for pre-worn jeans...

but they cost EXTRA, they are not sold at lower price after being preworn. Also, that is a fashion issue.

 

[Idontcare]

Originally posted by: dmens

Anyways, if it passes linpack then it is NOT necessarily working correctly with no errors... that is just one test out of a barrage of needed ones. I don't have the full list because I am not a CPU engineer.

reminds me of a funny story, one of the most unstable programs on prescott was the windows xp screensaver where the tetragon bounces around. it was so bad the thing got incorporated into production testing.

Yikes! Must have been hitting those double-pumped ALU's real hard to cause it to be a corner-case. I can't imagine the standard logic failing a screensaver computation while double-pumped ALU's had enough signal/noise to not be the weakest link.

 

[Fox5]

For people questioning whether or not the 4th core on a triple core is defective:

It could be no different than how cpu manufactures sell cpus at varying speeds. Many cpus are capable of speeds near the top of the line (and thus why overclocking works so well), but may not do it at the desired voltage. This triple core to quad core could be like that, just requiring a bit more voltage.

There's also a more insidious reason. AMD could have horribly planned their product mix. Phenom II X4's just aren't selling. They could do a broad price cut across the whole line, but that would cause a price war with Intel. It would also leave them with 0 chips selling for >$200. By artificially crippling the unsold X4's to X3's, they can sell something in a lower market segment (though ideally they'd want a sku specifically designed for the cheaper price point, but they don't have that) without slashing prices across the board.

This way, they may sell say 20% of their X4's at >$200, and sell the other 80% as X3's at $120, whereas the alternative may be to sell all the X4's at $130 and get less net revenue.

ATI and nvidia used to do this back in the day with graphics cards where they would sell their halo card, and then the best deal card was always the cut down version of the halo card. All the hardware was identical, but they'd artificially gimp it just to sell something in that price point. A classic example is the 9700 Pro and the 9500 Pro. The 9500 Pro had half its pixel pipelines disabled (they could be reenabled) and a cut down memory bus. Once the pixel pipelines were reenabled, it offered nearly all the performance of a 9700 pro for half the cost.

This would fit in with what I've thought about AMD's native quad core approach since they announced it though. Someone needs to get fired over their choice not to design a next gen dual core, or even a cheap quad core. (a quad core with no l3 cache would be substantially smaller, and they could have bumped up L2 cache to compensate, as well as not having L3 slow down memory access times...heck even a design with a smaller L3 cache might give them a better cost to performance ratio, but AMD apparently doesn't have the manpower/money to do even minor redesigns from their major server market skus)

 

[manimal]

nm

 

[Flipped Gazelle]

Originally posted by: Fox5

There's also a more insidious reason. AMD could have horribly planned their product mix. Phenom II X4's just aren't selling. They could do a broad price cut across the whole line, but that would cause a price war with Intel. It would also leave them with 0 chips selling for >$200. By artificially crippling the unsold X4's to X3's, they can sell something in a lower market segment (though ideally they'd want a sku specifically designed for the cheaper price point, but they don't have that) without slashing prices across the board.

This way, they may sell say 20% of their X4's at >$200, and sell the other 80% as X3's at $120, whereas the alternative may be to sell all the X4's at $130 and get less net revenue.

AMD can't be "crippling unsold X4's", because the only AM3 X4 available is the 810. The X3 710/720 are AM3 parts, whereas the X4 920/940 are AM2+.

It's a weird situation. Where are the X4 925/945? The 920/940 Phenoms are going to be discontinued in a few months.

BTW, a Newegg reviewer is claiming to have unlocked his 720BE to quad.

 

[richierich1212]

Well I wonder what mobo that Newegg reviewer has (or could be blowing smoke)

 

[Idontcare]

This XS link spells it out pretty clearly, basically you just need to set one option in the BIOS and bingo quad-core for the price of tri-core:

Mobo: Biostar TA790GX (they said you can use any 790GX + SB700 mobo to unlock)
Go to BIOS, modify Advance Clock Calibration to Auto
that's all
now save the BIOS, restart and wait to see the magic...

http://www.xtremesystems.org/f...howthread.php?t=218811

Lot's of screenshots there.

Looks like AMD has invented the poor-mans version of hyperthreading on a tri-core :laugh:

Here's one unlucky tri-core unlocker owner:

Even at 1GHz I get failures so not much point investigating further. More of a curiosity, for my little cpu at least. The memory controller is common to all cores I think, so there's not much point in playing with bus speed either.

It does boot at full speed, just means the bad core is not stressed during start up I guess. Here's the validation page

http://valid.canardpc.com/show_oc.php?id=510443

The other down side is that I can't use ACC, reports vary but some are getting an extra 200 to 300MHz with it enabled. It would be nice to hit the 4G mark.

http://forums.overclockers.co....wthread.php?t=17977875

 

[speedfreak]

wow! so it's a hit or miss thing when getting those X3? ACC is useless when you got a bad disabled core.

 

[Fox5]

Originally posted by: Flipped Gazelle

Originally posted by: Fox5

There's also a more insidious reason. AMD could have horribly planned their product mix. Phenom II X4's just aren't selling. They could do a broad price cut across the whole line, but that would cause a price war with Intel. It would also leave them with 0 chips selling for >$200. By artificially crippling the unsold X4's to X3's, they can sell something in a lower market segment (though ideally they'd want a sku specifically designed for the cheaper price point, but they don't have that) without slashing prices across the board.

This way, they may sell say 20% of their X4's at >$200, and sell the other 80% as X3's at $120, whereas the alternative may be to sell all the X4's at $130 and get less net revenue.

AMD can't be "crippling unsold X4's", because the only AM3 X4 available is the 810. The X3 710/720 are AM3 parts, whereas the X4 920/940 are AM2+.

It's a weird situation. Where are the X4 925/945? The 920/940 Phenoms are going to be discontinued in a few months.

BTW, a Newegg reviewer is claiming to have unlocked his 720BE to quad.

Hmm, I had thought all the Phenom IIs were AM3/AM2+. If AM3 is backwards compatible, what's the point in making two separate skus?
And the bulk of AMD's orders come from OEMs, not individuals like you and me. It could be that this X4 has an underwhelming OEM presence.

 

[Flipped Gazelle]

The 1st PhII's (AM2+) didn't have the DDR3-compatible memory controllers, whereas the AM3 PhII's can use both DDR2 & DDR3 RAM.

 

[blackpanther26]

When I first saw this on tom's I was like how can this happen? I mean I though once a company disabled a CPU core that there is no way to revive it.

 

[imported_Asianman]

its an interesting situation. I personally, don't think that this is a fake, I mean after all, not one in every two X4 phenom 2s are a failure in either a core or part of the cache(personal guess on the yield required). They shouldn't have permanently disabled the core, but since the people that tried it only activated ACC, its quite curious why AMD didn't test their own hardware on their own software first. And secondly, aren't the AM3 parts produced differently to the AM2+ phenom 2s? If they are, then this will have to be a fake since the only other AM3 part is a lower clocked X4 with a third of its L3 cache disabled. If I am correct, the fourth core shouldn't be doing any processing, but could still be monitored. Are there any figures available?
P.S, the picture was quite weird. It only showed the clocks, but not the processor model, so if its actually true, then the processor is an unlocked 920, an underclocked 940, or their equivalent AM3 models

 

[richierich1212]

Originally posted by: Asianman
P.S, the picture was quite weird. It only showed the clocks, but not the processor model, so if its actually true, then the processor is an unlocked 920, an underclocked 940, or their equivalent AM3 models

Take a look at the video I posted. It's weird, it turns into a "20".

 

[SunnyD]

Originally posted by: Flipped Gazelle
The 1st PhII's (AM2+) didn't have the DDR3-compatible memory controllers, whereas the AM3 PhII's can use both DDR2 & DDR3 RAM.

The first Phenom II's did have DDR3 controllers... in fact even the original Phenoms had the DDR3 controllers. They were just disabled, and the chips don't have AM3 pinouts.

 

[Idontcare]

Originally posted by: SunnyD
...in fact even the original Phenoms had the DDR3 controllers.

This is news to me, I had not realized this. Interesting.

Makes me all the more intrigued by the delay of introducing AM3/DDR3 Deneb's then if they've had an even longer learning cycle with DDR3 IMC than I was giving them credit for until now.