• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Zambezi vs Vishera, speed and # of RAM sticks?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

KetoSoi

Member
Joined
Jun 15, 2012
Heya guys,

I've been search'n a bit, and either I'm going (gone) crazy, or the topic I'm looking for isnt here anymore, lol.

I could have sworn there was a topic covering 2 sticks vs 4 sticks of RAM and their speed limitations (1600, 1866, 2133, etc) on the various cores AMD makes, particularly on Zambezi, with some info in the Vishera.

IIRC, for Zambezi OEM spec was 2@1866 or 4@1600... and I have no clue on Vishera. There was some discussion on how far people have gotten them to go, etc etc.

Anyone point me in the right direction?

Thanks :)
 
for Zambezi OEM spec was 2@1866 = correct and has not changed from Bulldozer to Piledriver that has been seen in AMD white papers that I am aware of. Never seen 4 @ DDR3 1600 mentioned by AMD. By the way AMD over the last 2 years or so had made viewing of their "real" whitepapers an NDA thing with what seems an actual need to know criteria as well.
 
for Zambezi OEM spec was 2@1866 = correct and has not changed from Bulldozer to Piledriver that has been seen in AMD white papers that I am aware of. Never seen 4 @ DDR3 1600 mentioned by AMD. By the way AMD over the last 2 years or so had made viewing of their "real" whitepapers an NDA thing with what seems an actual need to know criteria as well.

Thanks Rgone :)

I've been searching around the 'net, and there are several mentions of this limitation, but I'm not sure of the original source.

How much voltage have you seen thrown (safely) at the CPU_NB setting for the Zam' or Vish' processors?
 
How much voltage have you seen thrown (safely) at the CPU_NB setting for the Zam' or Vish' processors?

No idea what the DICE and LN2 boys use for CPU_NB but there are plenty of us on healthy water that run 1.4V a lot of the time. It does add pretty much to cpu heat of course but that is the price for high CPU_NB frequency. I know a few that use up to 1.5V CPU_NB a lot and have had nothing die.

In the AMD “Scorpius” Platform Technology – Performance Tuning Targets .pdf, AMD says that 1.35V to 1.45V is acceptable on air cooling for the CPU_NB. I just went and looked it up. But again it does add some heat to the cpu core temp. I think 1.18V is stock for most motherboard bioses on the CPU_NB.

RGone...
 
Yep I'm using 1.4vCPU_NB for the 2700 ish frequency when benching. My daily speed is a bit more reasonable with 1.325 @ 2500
 
the cpu/nb is good up to 1.55 I think I read, and have tried, but without the ram fo fill it 2500-2600 is all i can find reason to run.
 
On my current OC test config, I've got 4 sticks running at 1800. I've had to bump the CPU_NB voltage up a bit to get that stable. I've been able to go higher, but I've noticed that 2 temps in Asus Sensor, VCORE-1 and VCORE-2 start to climb rapidly, though CPU temp and Package temp (HWM) is still within normal range.

Trying to see how much I can get the 4 sticks to run at in relation to what the CPU will tolerate.
 
If you do a google search with the terms Asus Sensor, VCORE-1 and VCORE-2 start to climb rapidly, you will find plenty of users with similar problems. Most got some lower temps by having fans blowing on the VRM heat sinks to better remove heat from the VRM sinks themselves.

Honestly, most of us that stay here day in and day out and try to help have either the CHV or the Sabertooth 990 board and because of other reasons with software issues created by Ai Suite, most of us no longer will even use Ai Suite. So we don't spend any time looking at the VCORE-1 or 2 temps since those temps are only provided by proprietary Ai Suite.
RGone...
 
Hmmm, guess I'll rig up some sort of fan(s) for the VRM's then. This'll be interesting....

Thanks Rgone :)
 
the fans are a good Idea since you don't have a lot of airflow over that area. I'm just surprised the 4170 can work it that hard. I wouldn't have thunk it. The Sabertooth is a pretty solid board you'll like it.
 
No idea what the DICE and LN2 boys use for CPU_NB but there are plenty of us on healthy water that run 1.4V a lot of the time. It does add pretty much to cpu heat of course but that is the price for high CPU_NB frequency. I know a few that use up to 1.5V CPU_NB a lot and have had nothing die.

In the AMD “Scorpius” Platform Technology – Performance Tuning Targets .pdf, AMD says that 1.35V to 1.45V is acceptable on air cooling for the CPU_NB. I just went and looked it up. But again it does add some heat to the cpu core temp. I think 1.18V is stock for most motherboard bioses on the CPU_NB.

RGone...

I'll add that the Zambezi/BD chips (At least mine) can run over 3200+ with extreme cooling used (DICE in my case) and I had the CPU-NB voltage around the 1.6v mark BUT for anything else besides pushing it to the max with extreme cooling, I can't suggest doing that or that you'd even see 3000 on the NB - My 4100 typically hits around 2800 on chilled H2O using around 1.50v's, maybe a tad more than that but never beyond 1.55 since there is little to no gain by trying it and it only makes the chip run hotter.

The Visheras aren't as NB speed friendly, the max I've ever seen with my 4300 on DICE was around the 2850 mark, maybe a little over that and be stable enough to do a bench run. Again, I was using about 1.6v's to even hit that and the typical max I've noted with mine is about 2700 or so with chilled water being used and the same max voltage (Around 1.50v's) used. Of course some chips and setups might go higher if you are running good cooling for the chip.

On air, I'd be wary of going over 1.40v's for most any reason esp since what you get vs the chip's load temps, that can and will become a factor. I do agree that keeping the board's components cool helps and can get you further along but on my part, you'll have to try it with what you have, monitor load temps with the chip and see what happens but definitely be careful if you try it, no sense in killing the chip that way.

BTW I was also running just two sticks of RAM (4GB's total) when I did that, I honestly can't say if I tried it with 4 sticks used I'd be able to get it that high, DICE'd or not.
 
This is my current test config...

Running a slightly lower CPU clock than usual so I can eliminate it as a primary causal factor.

It runs quite well with this setup, load testing has no issue. I'd like to see how much more I can get out of the RAM, but the CPU Mem'cont seems to be the limiting piece.

I've ordered a Kingston RAM cooler to rig over the VRM. Should be here Mon or Tues.
 

Attachments

  • cpuz1.jpg
    cpuz1.jpg
    47.3 KB · Views: 369
  • cpuz2.jpg
    cpuz2.jpg
    31.2 KB · Views: 349
Just a little update...

In stalled a Kingston ram cooler over the VRM, and long story short, it lowered the VCORE temps 4-5*C. Much more than I thought it would get, I was thinking 2-3*C out the outside. Seems there may be some real benefit, at least on the Sabertooth R2.0 board.
 
Back