AMD "Piledriver" refresh of Zambezi - info, speculations, test, fans

[Dnomi]

Then show an hour of Prime 95 blend at 5GHz please.

I tryed, i cant whith Mugen 2 standard cooler anyway.. After a few minutes thread's stops working.

Maybee someone with a FX-8150, and better air cooling, will give it a try? Mine gets to hot:-P

[Zeus]

So not stable. I'm yet to see the first 5GHz Prime stable BD. Don't think it's gonna happen, not with water or aircooling.

[wez]

Nah you are probably right, dont think we'll ever see that on this gen.

Actually surprised there haven't been a bigger change in OC on launch chips and newer ones. On air/water at least. Not sure if that holds any truth at subzero.

[FlanK3r]

Im at 100% sure, there are some 5 GHz stable chips...If you have some vith good VID. Or 8GHz+ LN2 have great potential at air too (if you have not sh*t cooler). And you knows, PRIME is not the right stability tets for Sandy, Ivy and FX CPUs, there is many bugs and issues. Example with IB or SB you are OK in PRIME, but crashed in games or sometimes in idle! I have this experience not only personaly. The different problem is with FX. You got errors, but your CPU is really stable in others test. And you get this error at stock too, so...? No, PRIME is not good optimized last months and you must hardly to find right version of this software. And I have not time, so I prefer now simply AOD or OCCT. PRime was good one, two years ago, no problems, no issues with flase errors or issues with "real" stability in test.

[BeepBeep2]

Im at 100% sure, there are some 5 GHz stable chips...If you have some vith good VID. Or 8GHz+ LN2 have great potential at air too (if you have not sh*t cooler). And you knows, PRIME is not the right stability tets for Sandy, Ivy and FX CPUs, there is many bugs and issues. Example with IB or SB you are OK in PRIME, but crashed in games or sometimes in idle! I have this experience not only personaly. The different problem is with FX. You got errors, but your CPU is really stable in others test. And you get this error at stock too, so...? No, PRIME is not good optimized last months and you must hardly to find right version of this software. And I have not time, so I prefer now simply AOD or OCCT. PRime was good one, two years ago, no problems, no issues with flase errors or issues with "real" stability in test.

Most guys with 8 GHz + seem to note 1 GHz+ variance in worst to best core so I'd doubt it...

[Zeus]

well, we need something to gauge stability and i personally think Prime is still going strong. I know i could run Prime on my BD at 4.6+ and with more volts maybe 4.7+ but 5G was impossible. I haven't ever ran AOD myself but i'm pretty sure you won't find any BD cpu capable of doing that at 5G.

[hiepgia]

Trinity A10-5800K coming home!

[Dnomi]

So not stable. I'm yet to see the first 5GHz Prime stable BD. Don't think it's gonna happen, not with water or aircooling.

Why did you even mention prime, it's not stable with standart speed... You wantet to be sure my test failed?

Please note that im running cinebench stable, using a standart Mugen 2 cooler(Standart fan is quite, and that's what im using).. If i was to cool it better - A better air cooler would help, and/or two hi speed fans for push 'n' pull - No need to bother with water cooling.

[Daveburt714]

Actually surprised there haven't been a bigger change in OC on launch chips and newer ones. On air/water at least. Not sure if that holds any truth at subzero.

I find that strange too wez...
I picked up my 4th FX when MicroCenter dropped the 8120 to $139, thinking the process should've improved.

It's a decent chip (1210) and runs cooler, but my oldest 8120 (1135) is still the fastest chip I have (max clocks and pretty close stable) even though it is a heater!
Hopefully the new stepping (C0) 8350 will give us the same leap we got from PHI to PHII (I'm an optimist ).

Dnomi: I have 2 8120's and 2 8150's with pretty serious water cooling. None of them are daily stable at anything >4750...
Granted, stability depends on whether the apps YOU use run without crashing, but if you can't run Prime Custom(Max Mem) for at least 30 minutes sooner or later it's going to fail on you at the wrong time!

I'm not a big fan of running Prime for hours on end (it pushes all components close to the breaking point), but I'm not naive enough to say something's wrong with Prime as a stress test...

That being said, I wouldn't trust a rig that was Cinebench, IBT or OCCT stable if It wouldn't Prime for at least 30 minutes.

Looks like you have a nice chip, and if it serves your needs @5Ghz without crashing you've got yourself a golden sample!

[FlanK3r]

Trinity A10-5800K coming home!

nice,...looking at number of week, is very old chip... 17 week of 2012

[Zeus]

Not running Prime at default speed is completely new to me. If it was for you then something else must have been wrong or you were running the wrong version of Prime. I tested mine for at least half an hour and if it failed i had to increase Vcore Vcpu-nb or Vdimm.

I could run Cinebench at 5.15GHz yet stuggled to get it over 4.75GHz Prime blend. I guess it all boils down to what one calls stable. For me, like Dave, it's Prime or something alike.

[tbone8ty]

nice,...looking at number of week, is very old chip... 17 week of 2012

Yah wasn't there a rumor that they re spun these?

[FlanK3r]

Vishera will coming soon

And there is really the true sentence:
"AMD FX-8350 with Intel ahead slightly downloads, more revolution, but waits until chip architecture Steamroller, that will arrive sometime next fall."

[repman]

http://translate.googleusercontent.c...UGmhnK9Y0-Babw

[hiepgia]

My sample 8350 is arrived too.

Some first benchmark show it improved when compared with 8150. Good overclock. But did't test for power consumption.

[Alex-Ro]

Please do,it would be interesting to see.

[FlanK3r]

its better for OC than Zambezi or not? How good is Vishera in old x87 benchmarks?

[Alex-Ro]

I sense that flank3r heartrate is at 150 bpm in resting state now

[FlanK3r]

hehe, u right :-D

[fakhrain]

shouldn't it be called Zambezi C0 rather than Vishera, no?

would love to see A10 compared to C0 clock for clock in lightly threaded apps.

[Mechanical Man]

shouldn't it be called Zambezi C0 rather than Vishera, no?

would love to see A10 compared to C0 clock for clock in lightly threaded apps.

Yeah it looks like "certain person" might have had valid cpu to test 8350 performance.

That is real bad if it really is the case.

[informal]

Notice the C11.5 score of FX8350 on the top left . It's ~9pts at either 5Ghz or 4.8Ghz. At stock (4Ghz) it should therefore score : 9pts x 4Ghz /5Ghz=7.2pts if the 9pts was a score @ 5Ghz (which we don't know). For reference "blogger who shall not be named" reported the score on simulated FX8350 @ 4Ghz to be 6.89pts while FX8150 @ stock scores ~6pts. Therefore in best case scenario FX8350 will be ~1.2x faster than FX8150 or if "blogger's" result are closer to the real thing then 1.15x. The difference between two results(what is in the screenshot at presumably top OC of 5Ghz and "blogger's simulated FX8350) in C11.5 is exactly ~5% . There is a slight chance the top score of ~9pts is actually achieved at 4.8Ghz which will be too good to be true (IMO) or it may have been achieved at much higher (unstable OC) at ~5.5Ghz which might be the case.

[repman]

Notice the C11.5 score of FX8350 on the top left . It's ~9pts at either 5Ghz or 4.8Ghz. At stock (4Ghz) it should therefore score : 9pts x 4Ghz /5Ghz=7.2pts if the 9pts was a score @ 5Ghz (which we don't know). For reference "blogger who shall not be named" reported the score on simulated FX8350 @ 4Ghz to be 6.89pts while FX8150 @ stock scores ~6pts. Therefore in best case scenario FX8350 will be ~1.2x faster than FX8150 or if "blogger's" result are closer to the real thing then 1.15x. The difference between two results(what is in the screenshot at presumably top OC of 5Ghz and "blogger's simulated FX8350) in C11.5 is exactly ~5% . There is a slight chance the top score of ~9pts is actually achieved at 4.8Ghz which will be too good to be true (IMO) or it may have been achieved at much higher (unstable OC) at ~5.5Ghz which might be the case.

I noticed that the specs in CB say 4,82GHz on the left. Maybe that is from the last score?

[FlanK3r]

We know Info...FX-8150 need about 5500 MHz for this score... Good one .

[informal]

We know Info...FX-8150 need about 5500 MHz for this score... Good one .

Is it realistic that "Piledriver" in FX8350 really scores so high at "just" 4.8Ghz? Even at 5Ghz the result looks out of place to me. If it's 4.82Ghz like in the screenshot and the score was really 9.06pts than FX8350 at 5.5Ghz would score : 9.06x5.5/4.82=10.33pts ... Kind of unreal. This implies : 10.3/9=1.14 or around 14% "IPC" difference in Cinebench MT test. I'm not saying it's impossible but even Trinity doesn't score so high and it has the real PD core inside (not this tweaked Orochi core). Very strange results .

Maybe AMD scored 9pts at 5.2+Ghz but it was unstable so they repeated the test at more stable 4.8Ghz when the screenshot was taken. Notice the progression of the scores which is in line with continuous CPU clock scaling (like they OC bit by bit and reached highest score of 9pts at some point which may be 4.8Ghz but is quite unlikely IMO).