Orthos vs S&M/SuperPI
- Thread starter Dutch Guy
- Start date
SuperPI never was and still is'nt a stability test. S+M was written for the A64 platform and works very well with it. If you can pass 1 hour S+M on an A64, there is every chance it will pass Orthos/Prime for 8 Hours and more. On C2D and Intel Quads it's a waste of time, a much better replacement for it is OCCT, an hour or two of "mix" should give a very good indication of if Orthos/Prime will run 8 hours plus. An Orthos fail at 15mins running "Blend" stands a very good chance of being a ram problem, or if it's a 45nm cpu, the wrong GTL settings.
Is it difficult to explain what that is?
Yes
But i'll have a go.Most peeps seem to think that the GTL's and VTT just allow them to get a higher clock using less Vcore, and to a certain extent that's true. But they come into play much much more, with stability. To give a good example, look at the link in my sig. VTT at 1.17v and GTL's 68-67-69. I can run small fft's in Prime/Orthos all day long with values 3 or 4 either side of those, with no problem. But run large fft's or "blend" and it will fail within 10mins to 25mins. The only GTL values that work at my clock using 1.17v on the VTT are 68-67-69. Another thing that is a big problem with stability with 45nm's (and this applies to dual cores more than quads), is that peeps use much to much VTT in the first place because they are using it for the wrong reason (to get a higher clock with lower Vcore). In fact, and this applies on air as well as phase, the lower the VTT along with properly tuned GTL's, the more stable the clock.
Yes
Most peeps seem to think that the GTL's and VTT just allow them to get a higher clock using less Vcore, and to a certain extent that's true. But they come into play much much more, with stability. To give a good example, look at the link in my sig. VTT at 1.17v and GTL's 68-67-69. I can run small fft's in Prime/Orthos all day long with values 3 or 4 either side of those, with no problem. But run large fft's or "blend" and it will fail within 10mins to 25mins. The only GTL values that work at my clock using 1.17v on the VTT are 68-67-69. Another thing that is a big problem with stability with 45nm's (and this applies to dual cores more than quads), is that peeps use much to much VTT in the first place because they are using it for the wrong reason (to get a higher clock with lower Vcore). In fact, and this applies on air as well as phase, the lower the VTT along with properly tuned GTL's, the more stable the clock.
Did you find that you had to play with the VTT and GTL values much whilst you were trying to reach the clock in your sig? I only ask as I think mine are on 1.17v and 67% currently and my clock is both small fft and blend stable (8hrs+) and I'm just wondering if/when I'll need to tweak them.
Yes i did have to experiment a lot. On my DFI the values of 68-67-69 don't represent 67%, more like 63%-64%. The values are more critical when you are pushing the cpu to the absolute limit of stability, the further away from it you are, the less critical they become. My guess is that if you pushed your cpu speed up a lot higher than you are now, you would loose your current stability (Vcore aside), and would have to start Playing with your 67%. The mistake that you may be tempted to make, would be to raise VTT rather than alter your GTL values.
Last edited: