Newer Thuban's better than the older ones?

[theoldtimer]

Somehow my older Thuban got hurt (over volt in a beta bios I think). It lost a good 50-75MHz from its max benching O/C. This wasn’t on the CHIVF MB or any Asus MB as far as that goes. Anyway the old Thuban still runs fine it just lost a little of the top end.

That problem lead to this solution.



Although the better NB may be old hat to many of you guys it was a nice surprise to me. The NB clocked quite a bit better than my old Thuban right out of the box. I realize I haven’t tested it under any real load as yet but at no load it’s better, yes I know it only proves just that.

Default CPU volts are lower on this one also. All the bios voltage settings are @ default but LLC is enabled for both the CPU and NB. LLC took the CPU/NB voltage to 1.5v at the first NB multi increase.
It will be interesting to see it the new chip is better all the way around than my older chip? I’ll know in the next week or so? Hope I don’t hurt this one.

[FlanK3r]

3600 MHz NB?! whit air, watter? Post more, or show comparsion with older chip

[Titan7171]

sweet NB clocks there, that seems to be quite the improvement maybe 3.4ghz nb for 24/7?......mine runs 3ghz for 24/7 and this is a bad chip 3.2ghz on cold for benching. Good luck Im looking forward to your tests!

[Zeus]

That NB looks promising! Is it stable at that frequency?

[MrIcee]

Thats a great NB clock out of the box indeed, but look at the default volts: 1.5V. My 1013 chip defaulted at 1.35V and would do 3300Mhz out of the box. When running 3600-3700Mhz in competition I was pushing 1.55V so maybe there is a revelation here in sheeps clothing. Maybe, just maybe the CPU NB can handle more volts than initially thought. I certainly hope so, because I hate to admit what voltage I've had to run to hit 3900Mhz+ CPU NB under phase. But I will say this...my chip seems to have loosened up and OC's both overall CPU and NB much easier now after being tortured under high voltages and Mhz.

Great chip you picked up OT Can't wait to see your bench results down the road with that puppy

Best regards,

Randi

[demonkevy666]

you mind running 3dmark06 on that with stock NB and then higher NB ?

My 3dmark06 scores seemed low to me

[FlanK3r]

i dont tried max NB stability, but i can runing benchmarks as Cinebench with 3GHz and 1.22V at CPU-NB

[MrIcee]

Fellas,

As you probably already know, the key to unlocking performance in these puppies is the CPU NB. The higher you can run it with you're DDR3 maxed out, the higher the overall system bandwidth, which equals much better benchmark scores.

I'll also tell ya that these CPU's like and will run good memory at 2100Mhz+ with ease. That high DDR3 frequency coupled with as high a CPU NB as possible = killer performance. I posted up my last run in the June HWBot comp here, I'm going to find out soon how high I can push the memory on the CH4 and how much benefit there is. But I'll tell ya, performance was enough to plant me squarely between 4 LN2 benchers just running SS phase.

Best regards,

Randi

[zanzabar]

going from 2400 to 3ghz NB was worth .1s in super pi and windows boots alot faster, it was also worth a good 15% or so in everest across the memory

[crazydiamond]

wow nice oldtimer i was thinking i had a weak imc but only give it like 1.42v max i think time to do some more testing i guess

[theoldtimer]

3600 MHz NB?! whit air, watter? Post more, or show comparsion with older chip

This one is on chilled water FlanK3r. 1.368v is the default vcore for the old chip
This is the best NB clock I had with the old chip.

sweet NB clocks there, that seems to be quite the improvement maybe 3.4ghz nb for 24/7?......mine runs 3ghz for 24/7 and this is a bad chip 3.2ghz on cold for benching. Good luck Im looking forward to your tests!

Thanks Titan7171, this chip does appear to be better than my old one at least the NB.

That NB looks promising! Is it stable at that frequency?

No Zeus it isn’t stable enough to bench at that NB speed at least not at that voltage.

Thats a great NB clock out of the box indeed, but look at the default volts: 1.5V. My 1013 chip defaulted at 1.35V and would do 3300Mhz out of the box. When running 3600-3700Mhz in competition I was pushing 1.55V so maybe there is a revelation here in sheeps clothing. Maybe, just maybe the CPU NB can handle more volts than initially thought. I certainly hope so, because I hate to admit what voltage I've had to run to hit 3900Mhz+ CPU NB under phase. But I will say this...my chip seems to have loosened up and OC's both overall CPU and NB much easier now after being tortured under high voltages and Mhz.

Great chip you picked up OT Can't wait to see your bench results down the road with that puppy

Best regards,

Randi

Randi I left the bios set at auto for all voltage except VDRAM plus LLC was enabled for the CPU and CPU/NB. For the 3DMark06-NB tests posted below I disabled LLC for the CPU/NB. I thought maybe LLC was spiking the CPU/NB volts to 1.5v so that was the reason I disabled it. So with LLC disabled for the CPU/NB and the CPU/NB voltage set to auto the CPU/NB voltage indicated 1.125v as read in TurboV. That was at a NB speed of 2000MHz. When I clocked the NB to 3000MHz and up the bios auto fuction for the CPU/NB increased the CPU/NB voltage to 1.5v as before? I’m with you in wondering how much voltage these NB’s can handle? I’ve been taking it easy on the NB voltage on the old Thuban.

you mind running 3dmark06 on that with stock NB and then higher NB ?

My 3dmark06 scores seemed low to me

I’ll post those 3D06 tests after I answer everyone demonkevy666.

going from 2400 to 3ghz NB was worth .1s in super pi and windows boots alot faster, it was also worth a good 15% or so in everest across the memory

I haven’t tested those with this chip yet but I’ll get there sooner or later.

wow nice oldtimer i was thinking i had a weak imc but only give it like 1.42v max i think time to do some more testing i guess

Thanks crazydiamond, yeah I hear you all this time I’ve been trying to keep it under 1.45v.

[Daveburt714]

Very Nice OT!

I'm with you though, I'd be a little worried about 1.5v on the IMC...

I've done alot of testing with mine (probably too much ) and at least on my chip I don't see much improvement over 1.3v, then again I'm a voltage wimp (can't afford new gear right now... )....

I've only pushed mine to 1.35 and really saw mimimal return above 1.3.

Right now I'm stable @ 2950 with 1.3 and no LLC on the IMC.

Reguardless Gratz man!
Maybe you just got a golden sample, but it will be interesting as more new steppings get tested...

[Dumo]

Nice one OT

Imo with your chilled water, benching @ 3.7 NB and 4.7 cpu won't be a problem

My 1013s with "D" and yours with "E", probly your old 1090 with "C"

[theoldtimer]

Very Nice OT!

I'm with you though, I'd be a little worried about 1.5v on the IMC...

I've done alot of testing with mine (probably too much ) and at least on my chip I don't see much improvement over 1.3v, then again I'm a voltage wimp (can't afford new gear right now... )....

I've only pushed mine to 1.35 and really saw mimimal return above 1.3.

Right now I'm stable @ 2950 with 1.3 and no LLC on the IMC.

Reguardless Gratz man!
Maybe you just got a golden sample, but it will be interesting as more new steppings get tested...

Thanks Dave. This new chip (for lack of a better description) just has a different feel to it, kinda like my 555be I'll be getting a better handle on it over the coming days.

[chew*]

Batches make little diff ....

To many variables per chip nowadays.

Law of science are reflected in these chips.......it's all a give and take situation.

some chips have "balanced" cores, those will never be good for a validation.....Some chips are rather "unbalanced" those will never be great at multithreaded.......

Some chips have killer core speed......enter your first chip.....but at a price....weak NB.....Then there is the opposite spectrum.....strong NB.....but then weak Core speed.

Then there is another factor that throws a curveball.....how strong the mem controller is.....think becasue your chips has good NB it has a strong IMC?

Wrong.....thats another totally diff part of spectrum.

Best you can hope for is 2 out of 4, rare to have 3 out of 4.

For benching the best you can do is know your hardware and use whatever chips suits a particular benchmark best

[theoldtimer]

Nice one OT

Imo with your chilled water, benching @ 3.7 NB and 4.7 cpu won't be a problem

I hope so Dumo, I'll be playing real nice with this one until I find what she likes

[theoldtimer]

Batches make little diff ....

To many variables per chip nowadays.

Law of science are reflected in these chips.......it's all a give and take situation.

some chips have "balanced" cores, those will never be good for a validation.....Some chips are rather "unbalanced" those will never be great at multithreaded.......

Some chips have killer core speed......enter your first chip.....but at a price....weak NB.....Then there is the opposite spectrum.....strong NB.....but then weak Core speed.

Then there is another factor that throws a curveball.....how strong the mem controller is.....think becasue your chips has good NB it has a strong IMC?

Wrong.....thats another totally diff part of spectrum.

Best you can hope for is 2 out of 4, rare to have 3 out of 4.

For benching the best you can do is know your hardware and use whatever chips suits a particular benchmark best

One thing's for sure, the NB is the only part I've tested and not very much at that but it's only the beginning so now the fun begins.

[theoldtimer]

Here you go demonkevy666. The hardware is the same as in the first post. One used just one Asus 5870 GFX card at default clocks. The NB speed increase alone did help the scores a little.

3210.7MHz 16x200.7 NB=2006.7MHz


The score increased by 311 points but the CPU/NB voltage went up to 1.5v. (the voltage was set at auto in the bios+CPU/NB LLC disabled)

3210.8MHz 16x200.7 NB=3010MHz



The scores are still going up.

3210.8MHz 16x200.7 NB=3411.5MHz


I lost a little ground with this test even though the NB and memory speed is higher.

3203.7MHz 15.5x206.7 NB=3513.8MHz

[Dumo]

Probly your HT OT

[Daveburt714]

Probly your HT OT

Yeah, thats a serious HTLink drop for some reason....

The voltage spike on the IMC would consern me though, did you try setting it manually OT?

BTW: What do your chipset temps look like?
I've been doing some stability testing tonight myself and man my NB is getting warm (>63c)... I'm not sure it's a problem but I don't like it...

I wish EK would hurry up and release that mobo block for the CH4.

[MrIcee]

Yeah, thats a serious HTLink drop for some reason....

The voltage spike on the IMC would consern me though, did you try setting it manually OT?
.

Yep, the drop in HT is definitely the culprit there as its causing a bottleneck.

As far as the CPU NB voltage, it was set at auto and at that speed provided 1.5V, not too excessive and probably would get him some more NB Mhz.

Best regards,

Randi

[Daveburt714]

As far as the CPU NB voltage, it was set at auto and at that speed provided 1.5V, not too excessive and probably would get him some more NB Mhz.

Mine quits scaling way sooner than 1.5v, then again, I'm only on ambient water...

Not trying to hijack your thread OT, but what do your Chipset NB temps look like under load MrIcee?

[MrIcee]

Mine quits scaling way sooner than 1.5v, then again, I'm only on ambient water...

Not trying to hijack your thread OT, but what do your Chipset NB temps look like under load MrIcee?

Dave,

To be honest, I have no monitoring software installed: just a stripped OS and the bench I'm running. I can't even tell you my CPU temps LOL

Randi

[FlanK3r]

to CPU 3D06 score:
maybe its different between efectivity NB clock to cores clock. Think, every CPU clock has "ideal" NB performance clock. Example 3800 MHz/2700 MHz NB, 4000 MHz and 2950 MHz NB etc. And maybe better testing it in Cinebench R10 for it?

[Leeghoofd]

Why do you use AUTO voltages ? Have you tried with lower volts ?