Stomp stomp stomp!
- Thread starter Cob
- Start date
It's a Q6600 running at 'only' 3Ghz (my DS3 won't clock the quads over a 330mhz FSB. That will be fixed with a new mobo in the coming week or so).
And I'm running two copies of Ubuntu 7.04 in a pair of VMWare virtual machines on Vista x64, making a total of 4 linux SMP clients, one for each core.
But each frame is taking about 40-50% longer than it was with my e6600 at 3.6ghz (<30mins as opposed to <20 mins). Some of that difference can be attributed to the ~20% lower overclock, but it's still more than I'd hoped for. Vista isn't a contributing factor because I get the same times on x64 XP, tho VMWare Server v1.0.3 isn't yet fully optimised for Vista. Vista still isn't officially supported on the Server, tho the latest version of Workstation is fully supported and I'd expect the next release of Server to be tweaked a little.
So there's still plenty of tweaking left to be done, although I realise that I won't ever get double the production I was getting with the dual-core e6600.
And I'm running two copies of Ubuntu 7.04 in a pair of VMWare virtual machines on Vista x64, making a total of 4 linux SMP clients, one for each core.
But each frame is taking about 40-50% longer than it was with my e6600 at 3.6ghz (<30mins as opposed to <20 mins). Some of that difference can be attributed to the ~20% lower overclock, but it's still more than I'd hoped for. Vista isn't a contributing factor because I get the same times on x64 XP, tho VMWare Server v1.0.3 isn't yet fully optimised for Vista. Vista still isn't officially supported on the Server, tho the latest version of Workstation is fully supported and I'd expect the next release of Server to be tweaked a little.
So there's still plenty of tweaking left to be done, although I realise that I won't ever get double the production I was getting with the dual-core e6600.
Dropping to 3 clients drops all 4 cores to about 90%. It doesn't sound like much, but it's 40% of a core when you think about it. The three clients would crunch their WU's faster, but total output would be less than that crunched by four clients.
The best thing would be for VMWare to allocate more than two cores per VM. That way i could run only one VM and gain some performance. But sadly they only allow upto two cores per VM.
And myself and winSMP don't get on. I'm waiting for the next version before tearing out my hair again.
The best thing would be for VMWare to allocate more than two cores per VM. That way i could run only one VM and gain some performance. But sadly they only allow upto two cores per VM.
And myself and winSMP don't get on. I'm waiting for the next version before tearing out my hair again.
The couple of posts nearer the start of this thread kind of explain the difference. Tho I'm not completely sure why I only get 50% more points than I was getting with my e6600.
A quad will never give double the preformance of a dual-core based on the same dies. There's some diminishing returns going on. For example, both dies of the quad have to share the same resources as a single die dual-core has to itself. Similar to why SLi doesn't give double the preformance of a single graphics card.
I'm hoping to gain another 15-20% of my old e6600's output when I get a decent quad clocking board tho.
A quad will never give double the preformance of a dual-core based on the same dies. There's some diminishing returns going on. For example, both dies of the quad have to share the same resources as a single die dual-core has to itself. Similar to why SLi doesn't give double the preformance of a single graphics card.
I'm hoping to gain another 15-20% of my old e6600's output when I get a decent quad clocking board tho.