The first Venice tests are coming in, and the raw numbers are, frankly, disappointing. If you’re looking for 3GHz with air cooling, that isn’t happening.
However, this may not be the fault of the CPUs, at least in some cases.
One big problem with most of the results I’ve seen so far is that they either aren’t reporting the critical, potential bottlenecking settings, or if they do report some, the settings are aggressive.
There are three potential bottlenecks other than the CPU which can limit these Venice overclocks.
Memory speed: I don’t care what the people who sold you the RAM told you, if you’re running memory at more than 230MHz, that’s a potential bottleneck. You can’t say you know the true speed of your CPU until you’ve reduced memory speed down to around 200MHz.
Memory settings: Again, I don’t care what the RAM guys said, if you’re running fast memory settings, that could be what’s stopping you. If you want to determine the limits of your CPU, pull back on the settings.
Hypertransport Speed If you have Hypertransport set to 5X, odds are you’re not going to get very far, simply because by overclocking with that setting, you’re also overclocking the hypertransport links, and it doesn’t much like that. You won’t know the true max of your CPU until you lower it down to 3X (and, if things are going really well, maybe even 2.5X).
Of course, for all this, you need to lock the PCI/AGP/PCIExpress speed to default levels, otherwise, that could stop you, too.
It may sound odd to some that in order to go faster, you have to go slower, but this is something you have to at least test for to know the limits of your CPU. If you set your machine up with aggressive settings, and then say, “My machine only does 2.6GHz, this chip sucks,” you haven’t done the complete job. You don’t know the chip sucks simply because you haven’t eliminated the other possible causes. The chip may not suck at all. What may be doing the sucking is your settings.
It is only when you can say, “I tested this CPU while running the RAM at around 200MHz, with 3-4-4-8 settings and a 3X HT setting” that you’ve got the real job done. If the chip still doesn’t do more than 2.6GHz, then and only then will you know it’s the CPU.
But I Don’t Want To Lower My Performance!!!
Uhhh, a lower CPU speed reduces your performance, too, and generally, it has more impact on real performance than a slightly slower memory or HT setting.
Is that last statement always true? No, of course not, to say that’s always the case would be as untrue as saying that the most aggressive settings are always best.
There’s no simple answer to this, and anyone who says otherwise is not telling you the truth. Each situation is unique, and you have to experiment to see what mix of speed and settings work best for the particular components you’re using, and the software you use.
If you slow down all your settings, and find your CPU can only run 20MHz faster, odds are the slower speed/more aggressive setting mix will do best. On the other hand, if your CPU goes 200MHz faster, the opposite could well be the case. Either answer (and somewhere inbetween the two is probably what most will find) could be the correct answer for you, but the only way you’ll know is by going both ways.
Yes, “going both ways” is a phrase used in a much different context. Unfortunately, in some circles, if you suggest that slowing down some settings might be a net plus, they find that just about as well, outside-the-norm as some find, well, certain nondiscriminatory activities.
However, a real overclocker needs to explore all the options, that’s the real challenge to doing this.
Why Don’t You Do It?
That’s just what I plan on doing within the next two weeks.
In the meantime, though, if you’re going to jump ahead, in your testing, try going both ways. Try it fast, try it slow. If it doesn’t work out, no one will know unless you tell them, and who knows, you might like it. 🙂
For the benefit of others, though, it would be helpful if you did tell others how both ways worked (or didn’t) for you.