Sandy bridge OC Guide

[zalbard]

Does anyone know what's the maximum Intel-approved voltage on the SBs? Do they tolerate 1.55V that easily?

Unlikely. Just see current 32nm Gulftown and Clarkdale chips (1.4V) vs old 45nm Bloomfields (1.55V). I suppose it's around 1.4V for SB, too. They will eventually release a datasheet, though.

For Gulftown:



For Clarkdale:

[sin0822]

low cas 1600mhz kit like cas 7. or high speed 1866mhz 2133mhz+ RAM. Timings seem to have more of an impact on performance than speed unless teh difference is large. A quick way to calculate latency i always used was cas latency divided by overall speed. That was an old trick for DDR, but it might still spply to DDR3 as it seems to pre judge for you, but it doesn't take into accounts other timings or command rate. For instance cas 7/1600=0.004375 vs cas 8/1866=0.004287 then the 1866 would be better because of lower latency. But that might not hold true now, but if someone can confirm that woudl be nice.

[hiwa]

What would be the recommended RAM speed to buy for OCing SB? Is the GSkill 1600 set enough?

the 8gb and 16gb costs a lot right now just wait some days and u will find some nice memories from all manufactories, the IC are not defferent and we will see some OLD IC 2009 will back in action soon


very nice giude

[Durzel]

Can I ask a dumb question? It's been a long time since I've really tweaked with my overclock, and I'm still running a Core 2 Duo under a Vapochill LS which is going to be replaced by this board + 2600K. This guide seems really useful and I'm sure will ease my inevitable headaches with setting up my system. Whiilst being able to boot into Windows is a useful & simple way of diagnosing a bad overclock - surely just because it boots into Windows doesn't mean it can run Prime, or perhaps even run anything at all without bluescreening. What would be the process/flow chart for getting a stable Windows overclock?

[Boogerlad]

boot into windows and run your stress testing program
if bsod---> then lower clock speed
if successful---> enjoy

[Durzel]

boot into windows and run your stress testing program
if bsod---> then lower clock speed
if successful---> enjoy

But the guide doesn't talk about clock speed, just this turbo multiplier thing? (and vcore, etc)

[SoF]

sb doesn't really need an oc-guide when you are into the subject - took me like 1 hour getting used to the new platform and 2 hours to remark: still a lot of bios work to do...there will be some new things to discuss very soon and I hope there will be solutions...

[sin0822]

believe me as you don't need to use turbo, i don't and not many do.

[Leeghoofd]

Can I ask a dumb question? It's been a long time since I've really tweaked with my overclock, and I'm still running a Core 2 Duo under a Vapochill LS which is going to be replaced by this board + 2600K.

No need for vapochill on SB...

[Johnny87au]

No need for vapochill on SB...

Aircooling ftw!!

[Durzel]

Yeah Vapo is going to be ditched, no socket kit for SB anyway.

[Johnny87au]

Even watercooling would be utterly useless imo..

[sxs112]

breetai72: Multi-phase VRMs have never been "serial" systems. Additional phases are always in parallel so that each additional phase can supply additional current. Each phase is current limited based on the choice of FET and choke. The max current that a phase can support is about 25-30A with very good parts that are properly tuned. This design uses IR direct FETs which have excellent thermal dissipation properties, so I can see a 30% improvement in thermals, but nothing here would increase conductivity. The PWM drivers aren't anything new either. All in all, a good VR design, but nothing new.

Not really man, check the Volterra solution if you want to find a multi-phase VRM with serial connection between each phase. Also, the maximum current each phase can take may also vary with respect to the parts used. And yes, the increase in conductivity and things are just talking about component specs, just like what was talked about in sin0822's thread. (I do my research too, just couldn't be bothered spending the time to write them all out. After all, Asus only gave me a board to play with, and nothing more.)

breetai72: Nothing here helps latency. The best latency would be a single connection from the CPU to the x16 slot. Adding any switches or muxes adds latency. Even when the primary mux is set to route the CPU x8 to the primary slot (for a single x16) it adds additional latency.

Switches doesn’t add latency, Bridge does and that is the nf200. So routing to Passover nf200 when not needed is what is done here so you get pure native performance. If switches add latency, ALL X58 motherboards suffer then look at the PCIE lane switches on the Motherboards, yes even X58.

Etihtsarom: What would be the recommended RAM speed to buy for OCing SB? Is the GSkill 1600 set enough?

Anything that you want to use. Just remember that as the DRAM frequency is tight up with the BCLK, you will need to balance the DRAM frequency with your BCLK to find the best performance scenrio. i.e. (BCLK/100)*(2133 or 1600 or 1333)

Durzel: But the guide doesn't talk about clock speed, just this turbo multiplier thing? (and vcore, etc)

That's because under most cases the motherboard vendors handled the rest for you with their BIOS. In Sandybridge OC you can only play with the CPU Turbo Ratio and BCLK and various voltages anyway.

sin0822: believe me as you don't need to use turbo, i don't and not many do.

Turbo mode refers merely to Core ratio above the ‘stock’. Whenever you do that on i7/i5/i3, Turbo mode is enabled -> whether the BIOS of various vendors show it to be enabled or not.

[miahallen]

Not really man, check the Volterra solution if you want to find a multi-phase VRM with serial connection between each phase. Also, the maximum current each phase can take may also vary with respect to the parts used. And yes, the increase in conductivity and things are just talking about component specs, just like what was talked about in sin0822's thread.

Seems like a moot point to me if the CPU is the limiting factor

[Andrea deluxe]

can you explain the pciexpress repartition?

from the first to the last.....

nf200/cpu


please

[Olivon]

Nice guide sxs112, Thanks

[darckhart]

nice guide and preview thanks! i was slightly shocked at the high Vcore numbers you threw around, but i guess time will tell.

[sxs112]

can you explain the pciexpress repartition?

from the first to the last.....

nf200/cpu


please

[sin0822]

#1 what is a "q-switch" PCI-E switch?
#2 please explain the benefit and how the heck the NF200 quadrouples PCI-E bandwidth. It isn't great at doubling, but they wanna have it quadrouple. I would think SLI would be worse.
#3 Volterra???? are you sure? can you do me a favor and show me a volterra datasheet for the classy? or any real life final cpu implementation. FYI you can't just research Volterra, they don't post datasheets. You have to request them.

[bingo13]

cute guide, nice guide. hahaha massman, exactly then he tried to call me out, but I write everything, and i don't word for word marketing i do research.
This is nice, but I have a few points I would like to argue in the OCing section from personal experience, but its ok i will wait for release.

The issue is that you did not actually test the PWM solutions each board uses and then provide those results in a fair and unbiased manner. The main point of contention here is that you brought ASUS into an advertorial preview for Gigabyte and made statements/innuendos that are false. So yes, there will be responses to those false statements/assumptions and innuendos.

This post, the other user posts here and elsewhere for ASRock, MSI and other Gigabyte products did not directly mention or falsely state features about competing products. I have gone over this several times with you but in the interest of fairness we sent out a collection of boards to a respected third party to test your assumptions about the PWM setup on the UD7 and our board. As such, as I have stated in PMs, you will find your assumptions about our digital PWM setup on the P67 boards to be false along with a few surprises about your claims on the UD7. Do not say I did not warn you.

[bingo13]

No hard feelings .

I just wonder if good old Bingo13 will also chim in his opinion in this thread regarding the possibility that this is actually just ASUS marketing being re-phrased for forum purposes.

Or wait ...

Here is my chime...

This post and others like it for ASRock, MSI, and Gigabyte for new products are basically advertorials. Basically some marketing points that have been delivered by the various manufacturers and some personal experience by the user or users. I have no problems with it and everyone does it. Am I happy about it, not really as I think there is a different way to deliver the message.

The difference with the OP in the UD7 preview/review thread is that he presented marketing statements and/or false assumptions about features on our new boards without actually testing either our board or Gigabyte's board to confirm or deny those statements. As such, it is no longer an advertorial or preview but is considered to be a review. One that is fraught with inaccuracies that were pointed out by Peter and were discussed with the OP. However, he continued to push these false assumptions and so we will respond to them.

I have no issue if the "problems" he described about our new P67 boards were true, we are not perfect, nobody is, but I do have a problem when those statements/assumptions are outright false and the testing methodology is based on reading datasheets or reposting marketing slides. He made assumptions from documents and marketing slides without any scientific testing of the boards to confirm his (gigabyte's marketing message) statements. That is my problem and when it comes down to it, it should be yours also.

Honestly, if he had left ASUS out of this directly or changed his assumptions about our PWM/EFI setup based on sound test methodologies then there would not have been a response even though his assumptions about the capabilities of the Gigabyte PWM are also false in several areas. That eventually would have been brought to light by a reputable website or power user so no response from our end was even required.

I expect some negatives to be mentioned by reputable websites or power users about our boards. As I stated we are not perfect and will address those issues but at least we admit it. In fact we gave a wide variety of press members and enthusiasts the ability to test and comment on our products early on without strings attached. We had a lot of very positive comments and we had comments on areas of improvement that will be addressed in the next product cycle. At least we are open about our intentions and responses, no hidden agendas here unlike others.

[sin0822]

bingo13 glad you brough it into your own thread. Gary you never will back off, you tried to PM me my inbox was full i emptied it. I am asking your guy here questions becuase his write up is a ASUS marketing ploy, hey here is a Sandy Bridge OC guide that actually is an ASUS motherboard OC guide/review. Its ok because I have a real OC guide I made myself after a few weeks of testing and more than one chip. It has nothing to do with Gigabyte though. This guy restated ASUS marketing, do you know what I did? i wrote my article then saw the Gigabyte media kit and used them to back me up. Actually no one had anything good to say about what I said, no one said it was wrong either. Here is what happened.

Shamino asked questions I responded, he didn't like when I put down Volterra, becuase he pushed for it in the Classy. But now he works for ASUS so you tried to get him to make me look bad.

Next bingo13 you tried to make me look bad then you made threats, then you made more threats, how do you know I didn't take the board to a university I work at and have it tested? you really have no idea who i am man. But you are right I don't have your board, but I don't think it mattetrs unless Chil totally did a 180 and change digital PWM design. Chil better than Volterra, seems that isn't true, as far as most of the community is concerned, but you are saying transisent responce is better. Well this article says different: http://www.theoverclocker.com/backis...r-Issue-09.pdf

Then a last guy came into the thread and said the way i cam e to my conclusion was wrong, but that either way could be right, and that digital pwm transient responce might be faster by now, which is a real way to say hey there is nothing wrong but it might be different now even though its not i just had to say somthing to make you look bad.

Now you hate that article saying it was written by someone that works at interstill, you , ASUS marketing Manger, were corrected by a very nice member that he was an intern at the time.

Next we move on to does it matter? Hey if transient responce was that important, how come these boards aren't VRD 12 Certified? I see it no where in your marketing, but surly if Gigabyte has it written everywhere you would think it woudl be there for you guys.

Back on topic, "OC Guide" is short of true. you don't need to turn on C1E and EIST, you don't need to lower any of the other voltage, you don't call any OC Turbo short of confusing people. Next lets get to this, a boot into windows is NOT stable don't go raising multiplier. Any overclocker knows you have to test the stability, as yes this chip boots into windows it can run benchmarks, BUT try running Prime 95 or even better intel burn Test to test for stability.

One of three things will happen:
#1 processor will do the Stability test and pass
#2 processor is not do it and give you a blue screen before it is finished
#3 Processor will over heat becuase of its ability to exceed TDP and thus thermal throttle.

Most of these benchmarkers might wanna rerun their benchmarks, as I have a few aircoolers here and some just cannot handle the heat of these processors at load.

just somthing that really bothered me.

Also not every boot into windows is stable, most are 1/2 stable, aka able to run benchmark, but may throttle, or crash.

Can you explain to us why you want to keep on C1E and EIST?

Can you explain to us the question I had about the volterra data sheet?


Now back to my question about you guys taking 8x of lanes and making it x32, isn't this where the NF200 got its bad rep for increasing latency? Also does those switches increase latency?
I guess we gotta wait for a site to bench it against each other. Pity I also have my SLI setup on normal NF200 setup, I am just going to compare it against X58 SLI, if it loses then so be it.

[massman]

You know what, this retarded 'end-user' forum marketing. A while back I promised myself not to post on this forum anymore, or at least not to get involved in absolute useless debates.

I'll get back to the real OC forums like benchtec, OCA, classicplatforms and ocxtreme to discuss with real hw fanatics. So:

@ S.Bassiri: Nice thread, nice info.

@ SXS112: Nice thread, nice info.

At least we are open about our intentions and responses, no hidden agendas here unlike others.

No, you're not. The only thing you're open about is that you consider your own company the best out there.

[Gautam]

[Computurd]

You know what, this retarded 'end-user' forum marketing

dude i couldnt agree more with this. the rest about the other forums i dont agree with, but i do agree that these tactics are bull. this is marketing BS at its very finest.