Could B3 Phenom be the rebirth of AMD....+1000MHZ OC done!

[Tony]

Tony,

Why only 5-6-6 on the primary timings? Is that a chipset limitation?

It will do 5-5-5-15, this spd was to ensure at boot at default voltages. Plus the performance difference with phenom is tiny overall.

I found with this spd the lowest boot failure rate over the 4 enthusiast boards I have under test.

[Rammsteiner]

Still trying to get 2.7 GHz stable or have you pushed the CPUs up since then ...?

Still trying.

I started Prime Blend with afore mentioned Voltages and WinAmp running and I also downclocked the HT to 9x multi (so 1.8Ghz).

I come back and turns out the whole system had this super hardlock where USB devices are shutoff, screen black and power buttons dont work... This in less then 3 hours...

Temps are pretty good with the TRUE, so I dont think temps are an issue (sub 50's, around 47C or so). Ive the HT clocked at 2.2Ghz now... Wouldnt make sense though since a lot of people achieve better results with a lower HT, but well, Phenom is strange after all. HT @ 2Ghz made system freeze now and then, 1.8Ghz complete hardlock. Maybe 2.2Ghz increases things?

If not, maybe it's more a Voltage related issue, but as I said, I doubt that tbh.

[EDIT]lol, another post and another edit. Turns out HT multi's higher than default dont work on DFI. So testing a little higher bit of CPU, 2760. But both NB and HT at 2.4Ghz though. If this crashes, I actually wouldnt care because I want to see HOW it crashes; does it crash as the last one or more a freeze. But in the end I hope it doesnt of course.

[EDIT 2]Okay... My system didnt crash a complete 720 degree loop... Im just ignorant and will try the HT multi at 7 again then.... Holy , for the first time I was worried I killed something lol. Also Ive the feeling thrwoing more Voltage dont really help a lot... Maybe previous posts about an 'optimum temp' for the Phenom is true-ish, or Voltage, I dont know. ARGH, SOMEONE GEIV MANUAL NOAW!!111 lol. It would even be easier to get a whore for free than get a Phenom run nice

[KTE]

Here's someone who hit 3.415G benchable with 9850BE and compared it to QX9650 at nearly exact MHz in 2k6: http://www.octeamdenmark.com/forums/...30&postcount=1

Vista Ultimate
Nv 8800 Ultra

CPU: 9850 BE - QX9650
Score: 13,815 - 13,786

But I believe most of this is already well known

So basically, the perf. difference between Penryn<->X2, a colossal deficit, is now no where near there, its a win some lose some situation between the two now clock for clock, but the main problems being: no high clocks and higher than 45nm chip power.

[biohead]

Here's someone who hit 3.415G benchable with 9850BE and compared it to QX9650 at nearly exact MHz in 2k6: http://www.octeamdenmark.com/forums/...30&postcount=1

Vista Ultimate
Nv 8800 Ultra

CPU: 9850 BE - QX9650
Score: 13,815 - 13,786

But I believe most of this is already well known

So basically, the perf. difference between Penryn<->X2, a colossal deficit, is now no where near there, its a win some lose some situation between the two now clock for clock, but the main problems being: no high clocks and higher than 45nm chip power.

http://www.xtremesystems.org/forums/...&postcount=282

[Rammsteiner]

Yeah KTE. Yields, and then TRUSTABLE yields aren't still there sadly.

But when B4 (?) hits, or the 45nm's I think and hope we'll see better clocks.

Stramge, higher Vcore I got almost same temps although lower Vcore seems to be more stable. Both for Vnb and Vcore btw. Running now HT at 7x. Ill leave Prime 95 belnd on trough the night, same with a minimized EvE-online account, 2x CPU-Z, 1x Core Temp and WinAmp (speakers off).

Lets see what that brings stability wise. Also Im wondering, since HT is actually the link between all add-ons and CPU, what's the lower limit you dont want to get in? At some point you would bottleneck the system I guess.

And what would happen if the HT runs at 1.4Ghz, but you OC the PCI-e bus to 105~110Mhz? How does this reflect in bandwith?

[Loser777]

I sure hope 45nm breaks some WRs... Anyone tried 9850 on LN2 yet?

[KTE]

http://www.xtremesystems.org/forums/...&postcount=282

Oops, I totally forgot those were his exact scores, apologies.
You can always see another 3.3G with 1.35v: http://xtreview.com/addcomment-id-48...potential.html

And what would happen if the HT runs at 1.4Ghz, but you OC the PCI-e bus to 105~110Mhz? How does this reflect in bandwith?

Never tried it.. you tel me?

[QuietIce]

[snip] Lets see what that brings stability wise. Also Im wondering, since HT is actually the link between all add-ons and CPU, what's the lower limit you dont want to get in? At some point you would bottleneck the system I guess.

And what would happen if the HT runs at 1.4Ghz, but you OC the PCI-e bus to 105~110Mhz? How does this reflect in bandwith?

A lot of that would have to do with the load from other components - especially the video load. Tony posted a chart a little ways up showing the different 3DMark scores for increasing HT Bus speeds on his rig and it was still climbing when he hit 2000 MHz. But if you don't have dual X2 graphics it probably wouldn't bottle-neck at all near top end.

I'd say you're down to comparing 3DMark scores for each rig at various HT Bus speeds. Just remember most K8's ran 1000 MHz stock. I've never seen a DX9 card top that out - not even in SLI. I'm sure DX10s can though ...

[aGeoM]

Multi's - I change using Sam's MSR Tool, editing 0xc0010071, 0xc0010070, 0xc0010068, 0xc0010067, 0xc0010066, 0xc0010065, 0xc0010064, 0xc0010062.

eg.

Current setting: 200x14x10x10 = 2800MHz, 4 cores, 2000MHz NB/HT

[MSR1]
index=0xc0010070
high=0x00000000
low=0x2800280C
mask_low=0xFFFFFFFF
mask_high=0xFFFFFFFF
coremask=0b1111

[MSR2]
index=0xc0010064
high=0x800001EE
low=0x2800280C
mask_low=0xFFFFFFFF
mask_high=0xFFFFFFFF
coremask=0b1111

[MSR3]
index=0xc0010065
high=0x800001B0
low=0x25009080
mask_low=0xFFFFFFFF
mask_high=0xFFFFFFFF
coremask=0b1111

[MSR4]
index=0xc0010062
high=0x00000000
low=0x00000000
mask_low=0xFFFFFFFF
mask_high=0xFFFFFFFF
coremask=0b0000

Now edited to give: 200x14x13.5x14x13.5x10x10, 2800/2700/2800/2700, 2000MHz HT/NB

[MSR1]
index=0xc0010070
high=0x00000000
low=0x2800280B
mask_low=0xFFFFFF00
mask_high=0xFFFFFFFF
coremask=0b0101

[MSR2]
index=0xc0010064
high=0x00000000
low=0x2800280B
mask_low=0xFFFFFF00
mask_high=0xFFFFFFFF
coremask=0b0101

Thanks Tye, didn't work for me, nothing changed, do I need to reboot?, just pick up my 9850BE 0809EPMW , settings in bios: CPU-FID=14, HTT=200, HT=2000, NB=2400.

[KTE]

^ It will change them, something must be going wrong, most likely you just need to refresh - registers are where the CPU gets the current operating info from-> Give me a list of what settings you're using for all of them, i.e. the hex values.

Enable CnQ in BIOS, make sure it works, get into windows and choose the highest frequency mode (i.e. disabled CnQ), change the correct parameters, run the exe with your new settings, then change power modes to enable CnQ [active] and then revert to the highest frequency mode you started in back again. You'll see the new settings.

i.e.

[Particle]

Here's someone who hit 3.415G benchable with 9850BE and compared it to QX9650 at nearly exact MHz in 2k6: http://www.octeamdenmark.com/forums/...30&postcount=1

Vista Ultimate
Nv 8800 Ultra

CPU: 9850 BE - QX9650
Score: 13,815 - 13,786

But I believe most of this is already well known

So basically, the perf. difference between Penryn<->X2, a colossal deficit, is now no where near there, its a win some lose some situation between the two now clock for clock, but the main problems being: no high clocks and higher than 45nm chip power.

From the page you linked to:
"...and I havent seen a 9850 that couldnt do 3000 or more mhz, yet. Most of them does 3200-3400mhz at arround 1.4v..."

I'm not sure where this guy has been. Of the results I've seen posted, easily half won't make 3GHz stable. If you look at our own poll or the OC database, you'll see that 5 9850s are stable < 3GHz while only 2 are above.

[QuietIce]

I'm just letting the rig crunch for now @ 3008 MHz and waiting on WC.

From the page you linked to:
"...and I havent seen a 9850 that couldnt do 3000 or more mhz, yet. Most of them does 3200-3400mhz at arround 1.4v..."

I'm not sure where this guy has been. Of the results I've seen posted, easily half won't make 3GHz stable. If you look at our own poll or the OC database, you'll see that 5 9850s are stable < 3GHz while only 2 are above.

Make that three rigs above 3.0 GHz - mine hasn't been added yet ...!

[Particle]

Congrats!

---

My point of course being that it isn't hard to see a single 9850 not making 3GHz. This guy's comments are going to be misleading to whoever reads them.

[QuietIce]

My point of course being that it isn't hard to see a single 9850 not making 3GHz. This guy's comments are going to be misleading to whoever reads them.

It may not be the chip so much as the other hardware (MB, RAM, VC) and us. It's a new platform and there's still a lot to learn about hardware compatibility and OC'ing this beast. IMO it's more a matter of acquiring the knowledge to put together the right system and adjust the settings correctly than anything else. I lucked out on my OC, just look at the odd settings I run to keep it there - 188x16! There's got to be a better answer than that!

Eventually we'll figure out the voltages, multipliers, and all that to get these chips OC'ing well. We'll learn that xxx RAM doesn't work so well on that board or this VC needs a little more (or maybe less) voltage than normal. But all that will take time and lots of experimentation just like any other new platform ...

[jonspd]

that guy knows more then you and I put together about amd Phenom chips IMHO


however not everyone will be able to run 3.4ghz as you stated but then again who's to say that it want end up getting there by the end of this reversion or with some newer board's?


just read alittle bit of his information in this 71 page thread http://www.xtremesystems.org/forums/...d.php?t=167447

[Particle]

I guess I just have to disagree with your interpretation. To me, the question I raised versus what he said was very cut and dry. He says one thing while reality shows another. His knowledge about a certain subject, where we might get to eventually, or our inability to maximize a Phenom's potential are largely irrelevant to the statement made. It is possible that he wasn't exaggerating at all either, if he genuinely hasn't seen how most (and I say that factually, not stereotypically) of us aren't achieving what is suggested by his statement. I guess I just don't see why an obvious exaggeration would be so readily defended by others who have seen the same data I have. :/

[QuietIce]

I can certainly see your point, it's not that. Obviously no one can just sit down and be running 3.0 GHz stable in a day at this point. But when I see things like people running 3.0 GHz seemingly stable for several days then getting a freeze I have to wonder if it's more of a setting or hardware compatibility issue (or even BIOS) than the CPU - you know what I mean?

But maybe you're right - it could be these chips are inherently more unstable than what we've worked with in the past. Only time will tell ...

[KTE]

From the page you linked to:
"...and I havent seen a 9850 that couldnt do 3000 or more mhz, yet. Most of them does 3200-3400mhz at arround 1.4v..."

I'm not sure where this guy has been. Of the results I've seen posted, easily half won't make 3GHz stable. If you look at our own poll or the OC database, you'll see that 5 9850s are stable < 3GHz while only 2 are above.

That guy you're referring to I don't know, it was a news posting on that site. But you have to realize, no where did he mention stable and he probably did not expect the weird idling issues we're seeing either. There are quite a few 9850 that do 3GHz and over load stable, which any normal user would expect is fully stable, but with Phenom that's not the case. Just the fact that Phenom B3, the same Phenom a few months back which could hardly hit 2.55G stable is now hitting these frequencies is a big change for Phenom itself, and from what I understand by the linked post, that is what he's referring to; B3 being physically able to hit those MHz, which was not possible before. Mine is also doing 3055 1.35v load stable but no where near that for general use.

I guess I just have to disagree with your interpretation. To me, the question I raised versus what he said was very cut and dry. He says one thing while reality shows another. His knowledge about a certain subject, where we might get to eventually, or our inability to maximize a Phenom's potential are largely irrelevant to the statement made. It is possible that he wasn't exaggerating at all either, if he genuinely hasn't seen how most (and I say that factually, not stereotypically) of us aren't achieving what is suggested by his statement. I guess I just don't see why an obvious exaggeration would be so readily defended by others who have seen the same data I have. :/

Jon thought you were saying I was saying that most likely
There is no defense of <and so on>, what you said is correct to a point and what the linked post said is also correct to a point. The difference being what you are talking about is full stability and what he was saying "looks" to be in regards to just physically being able to hit those MHz. Trust me, I tried a few Phenoms and none did hit these MHz as easy as this before B3 on stock coolers.

[Rammsteiner]

There's a Phenom 9750 for sale in here which has 95W TDP and 4MB L2 cache, anyone heard about that? That's quite interesting, new stepping coming up already?. Or is this just a better binned chip and the shop messed up the 4MB L2 cache with L3 including? Although all the other Phenoms show the real 2MB L2 cache...

Since the shop doesnt work with links, www.sallandautomatisering.nl, on the left under 'PC Onderdelen' go to 'Processoren', 'AMD', 'AM2', there on the bottom you'll see the Phenom Im talking about...

Interesting nonetheless, no matter if they accidentally messed up the L2 cache, the 95W TDP is....

[Brother Esau]

Update:......Hitachi Drive Fitness Test does indeed work on the SB600 Controller but You Must download the Latest Build and use the CD ISO Image and remeber to enable USB Mouse Support in Bios.

[BertM]

There's a Phenom 9750 for sale in here which has 95W TDP and 4MB L2 cache, anyone heard about that? That's quite interesting, new stepping coming up already?. Or is this just a better binned chip and the shop messed up the 4MB L2 cache with L3 including? Although all the other Phenoms show the real 2MB L2 cache...

Since the shop doesnt work with links, www.sallandautomatisering.nl, on the left under 'PC Onderdelen' go to 'Processoren', 'AMD', 'AM2', there on the bottom you'll see the Phenom Im talking about...

Interesting nonetheless, no matter if they accidentally messed up the L2 cache, the 95W TDP is....

i found the link:

http://www.sallandautomatisering.nl/....php?pid=48682

must be a mistake because its not mentioned on amd site:
http://www.amd.com/us-en/Processors/...E15347,00.html

[Rammsteiner]

Hmmm, I always get lost on AMD sites and never find these things lol.

Ty though. 4MB is an error then, but the 95W TDP is correct... How would those OC.

Im gonna save some cash up and get watercooling, the new 790FX/SB750 board and the HD4870's lol. And might get another Phenom if this one turns out to be a bad chip with the fixed 790FX.

[Oc-Ghost]

I always use www.amdcompare.com
and about 9750 125/95w difference between the two is
voltage((125/95) 1.20/1.25/1.30V / 1.10/1.15/1.20/1.25V) notice 3 vs 4 modes ,
and max temp((125/95) 61/70 degree C).

[Rammsteiner]

I always use www.amdcompare.com
and about 9750 125/95w difference between the two is
voltage((125/95) 1.20/1.25/1.30V / 1.10/1.15/1.20/1.25V) notice 3 vs 4 modes ,
and max temp((125/95) 61/70 degree C).

Oh... tempting...

[biohead]

Oh... tempting...

yah, higher max tcase = higher air OC generally.