After a bit of drama I recently managed to convince my bosses we needed some new hardware in the office - and after a little more drama talked them out of going default-Dell and "gamble" on an independent AEC-dedicated workstation/server provider who destroyed Dell both on spec. matching and price matching.

So on Monday 6 nex boxes arrived, all running an i7 950 @3.06Ghz, 12GB RAM and a 2GB Ati V7800, which (on paper) should be quite a step-up from our Dell Precision 390s running a C2D 6700 @2.66 , 8GB RAM and a 256MB Ati V7200. But the thing is, running an identical render side by side on the two machines "only" reports a 30% improvement in output time.

Maybe I'm expecting more than I should - and certainly we didn't build the machines to be render monsters - but in day-to-day working we're already finding there's been little gain in the upgrade.

The thing is, I have my suspicions about our IT-control policy. Our director swears he's not got the IT guys loading bloaty monitoriing software (despite silly startup times) but regularly champions the use of a generic disc image for builds, (regardless of hardware) and then uses Revit as a scapegoat for all IT-related crashes & errors.

Soooo.... what synthetic tests are out there that I might freely download and use to test the out and out system performance of these machines? Ideally I want something that'll assess things fairly quickly (we can't have downtime). Beyond that, what other way means are there for me to see if all things are hunky dory?

I'd expect more than a 30% improvement myself.

You could try running folding@home on them to see the difference in performance

Cinebench will do it as well http://www.maxon.net/downloads/cinebenc ... h-115.html

both these are great, might also be worth giving the processors a test with Prime95 and IntelBurnTest (also works on AMD CPUs ), theres a list of IBT results here so you can see where your system is performing relative to others with the same CPU

GPU tests are a little harder to come by, especially ones that dont require installation. i think FurMark can be extracted from the installer and run standalone, but i havent tried myself

+1

I had a C2D laptop and a C2Q desktop and I got an i7 QM laptop and it pinkled all over both of them.

PC Pro also did a rendering test program a few months back (June/July?). My C2D managed it in around 4 minutes, the Pentium Ds that the company were buying for their power users ran it in around 12 minutes and my laptop in around 2 minutes...

http://www.pcpro.co.uk/blogs/2010/07/30 ... rkstation/

I would expect a Core i7 in a domain, with AV software, to boot in well under a minute, loading the profile depends on the amount of crap on the desktop and network performance, but 20 seconds or so should normally be enough.

If it takes much longer, then they have seriously botched something up. As an Admin, I'd be ashamed of myself, if the boot time took more than 2 minutes to desktop on any half way modern machine.

As far I know : our AV is run server side, we have a locally hosted exchange, run gigabit lan on cat 5, have no roaming profiles, nothing configured (at least to show) in window's "startup" folder, a do-not-work-from-desktop QA policy (but most of us have loads of shortcuts to an array of mapped network drives), and we only connect to our network license (virtual software based) servers on request... but still startups do take considerably longer than 20 seconds!?

I can fire my machine up and still get out of my jacket and say hello to a few people before it gets to login, after which I can go make a coffee and/or go to the loo before my desktops settled enough for me to start asking to open outlook, etc. We've quite a few scripts and "group policy" items that flash up on startup - it's these I have serious doubts about.

I was also horrified to find that the disc image used didn't install an audio drivers so none of the users can watch online tutorials that I send them (in the first instance our office manager was instructed to respond to their requests with "they don't have soundcards" which is kind of correct, but our mobos have onboard 7.1, so its obviously a concious omission made by our director that I'm challenging). Also, and this is far worse, discovered the paging file was default-set to 2046MB, which I had to manuallly change to the recommended 18426MB. My director reckons this wasn't an IT oversight in making the disc image - but I'm contesting I don't care about the image, I care about the finished article that's been wired up for use (ie. they're not ready). Unsurprisingly since increasing the paging file, things have improved, but still something feels up.

Our director reckons it might be the network, but is doing little in the way of troubleshooting it, and since we're not allowed to work locally, we're just being told in effect to "sit and suffer" - which is an absolute farce in a time-poor industry such as architecture if you ask me.

Beyond benchmarking (which only reports the performance, rather than offers ideas in the way of improving it) how would you lot go about really assessing whats up?

Get in an external company to do a network audit, they'll find something for sure.

Why don't you offer, Sas?

Because that would be against forum rules

Sorry, my post was intended as a vaguely humorous workaround of said rules. Was trying to be helpful. I'll go away. Sorry.

This might be a red herring in your case, but there are some workloads (such as Pro/Engineer's annoyingly single threaded model regeneration) where the large L2 cache on older Core2 processors can prove an advantage and make comparisons with otherwise superior I7 processors more even than you might expect.

Have they built the disk image around the new motherboards, or have they just used the old image / modified the old image? It could be a driver conflict problem, or they don't load the right drivers for disks, processor, north/south bridge etc.

"apparently" they rebuilt the image, but I'm betting they modifed an old one.

"thankfully" yesterday saw a massive drop in network performance for everyone, across all applications, which spurred one of the other directors to finally sit up and make a noise themselves. Methinks on Monday some real investigation might actually start.

Only get a network audit done if you know there's a problem with the network, or you don't know what state it's in!

We were recently pressured into getting a network audit done - it's the most useless document I've ever read. In our office we don't have enough floor ports, so we have a 5 or 8 port switch under almost every cluster of desks in the building. Very very far from ideal, but we can't get the money to get more structured cabling done (and believe me, we want it done more than anyone!).

Certain important people have been complaining, so an audit was done. Their suggestions? Get more cabling done, and replace a load of our switches with a beefy modular beauty. Exactly what we have been asking for. Oh, and they suggested changing our naming convention. Brilliant.

Well that's the thing - I've quite a bit of faith in our IT support - it's whats asked of them that I question. But like I said, if things continue as they have, something will have to be done - you simply can't have an office (that relies entirely upon computer based information and workflows) running on a shabby network.

As was mine