Higher OC vs More RAM

AaronD

AaronD

AaronD

Associate
Joined
6 Jan 2010
Posts
43
Hi

We have a number of workstations that need lots of processing power. The Autodesk products we use aren't that multi-core compatible and it's the processes that only utilise 1 core that we want to improve. So no point in multi socket xeons etc.

So I'm looking at OC'd i7s

Two questions:
1. Would be be better off with a much higher OC than with 12GB RAM for example (the processes in question use little RAM so a bit of a no brainer really... I'll ask anyway :))
2. How do the high OC'd i7s compete against something like a Xeon W3530 2.8Ghz out of like a £3.5k machine?

Thanks
Aaron
 
Aaron to give any kind of informed answer a bit more information would be good.

Just how much ram is needed to run the application?
How many other applications will there be running at the same time?
How critical is the information or the work that is being done? The reason for this question is that inherently ANY over clock is potentially unstable. If you have mission critical processes going on then I would strongly suggest not using a over clocked system. And trust me mate I know what I am talking about when it comes to over-clocking.

That being said.

1 If you do not need the ram do not use it. More ram just creates more work for the IMC to deal with.
2. The SB i7 processors clean up the competition on single threaded applications with or without a over clock.

The reason that Xeon processors are slightly lower clocked is for the dependability of the server environment where they are designed to be used.
 
cjgarneds - xeons are actually used in workstations too, in business that is.

1Day thanks for response

The application RAM requirements (take autodesk revit for example) are actually very varied due to the nature of the application / project. Put it this way, we could happily run along on 4GB doing day-to-day stuff.

How many other applications, really just outlook and your typical background apps / services.

We are willing to compromise on stability, though we actually have an i7 @ 4.2 for quite a while and it's been rock solid. Hence my curiosity.

What's good about OC'd machines we find is that we can have a much higher turnover / shorter lifecycle rather than purchasing top spec machines and keeping them for 3+ years. We can actually keep up with the very fast pace of the software development and project size.

To give you a little more background info. Say we have 10 machines. Those 10 are fine for the general day to day 3D work that is required. But there are times we have to import / render stuff and that's where the performance requirements shoots through the roof, seriously. We have one thing at the moment which is a bunch of co-ordinates in a CSV file at over 250,000 lines long. It takes forever. And importing is not multi-core compatible... This only needs to be done once for a job, then back to usual day-to-day stuff.

We will not be replacing machines, merely introducing, a couple at most, highly OC'd machines just to run the intensive procedures that are only done every so often. Thus stability, 24/7 usage etc is not an issue. Due to the cost, we can then just throw them in the bin and replace say every 12-18 months. You couldn't do that with £3.5k+ machines :D

With me, or silly idea? :D

Your 2nd point is interesting.
 
I hear you Aron, based on that information then yes you can certainly do what you are trying to achieve. There are ways to make your system as fail proof as humanly possible and other little tricks that can improve through put of the nature you are looking for. After all that is exactly what I do as a bencher all the time.

Last question. How big are the files in MB or GB that you would like to work with at this hyper warp speed. :)
 
Thanks, glad I'm not going mad then.
Will find out for you, will be popping into the office shortly.

With a score like that you should hire yourself out to people, for overclocking expertise that is :D. Damn. That would save us so much time. To put into perspective that 250,000 line csv import takes over 2 hours!
 
I would suggest that you use two C-300 128GB in RAID 0 for your system. The SSD will greatly speed up the process. I suspected at the beginning of this discussion that it was more a bandwidth issue rather than a purely processing issue. The ram capacity used during modelling can be checked by using the system monitor function. You might be surprised at how much ram is actually utilised.
 
You must log in or register to reply here.
Back
Top Bottom