What do you want reviewed?

CM Silencio 550 review would be nice. Just got on sale in around here and seems like a good competition to the Fractal design case

 

The whole point of the rep system is that anyone is free to give rep as and when they feel it necessary. To review every post and decide if it can be repped would be the only way to cut out your gripe - old, useful threads would be excluded from rep otherwise.

[edit] That sounded a bit strongly worded, it wasn't really supposed to be having a dig.

 

This hit our lab late last week. Currently the plan is to get it into the next issue (96) of the mag and on the site either just before or just after, depending on the site schedule, so stay tuned.

 

Sweet, thx

 

I have a request for a review, but it could be time consuming.

Working as an engineer and being a computer enthusiast, I'm sometimes asked to build PCs that can do CAD and/or simulation work. Mainly this is to be done as cheap as possible and hence I end up using a Geforce card (AMD drivers SUCK in Linux) but a couple of systems I've been involved in have had larger budgets. In these cases, I've confessed that on the GPU side of things I don't know what would be better for a system, an expensive Geforce or equivalently priced Quadro. One system ended up getting a £1000 Quadro just to be safe since the fellow running it was reconstructing MRI scans.

Anyway, what I'd like to see is a comparison for Geforce and Quadro card running various CAD programmes, particualarly paying attention to equavalently priced versions. Things like:

Anti-aliasing and image quality
Frames per second
Stabilty (my GPUs can end up needing a restart after a day's work using CAD)
Comparison of OpenGL vs standard rendering
Memory usage of cards
And any other things you can think of.

Cheers for reading!

 

That's a good suggestion Pete J, and something I've been thinking about the last few months.

The principle problem is finding realistic benchmarks for professional cards, as getting hold of the apps and even more importantly proper 3D models/simulations is nigh on impossible as they are very valulable IP to the creators.

What apps/tasks do you, or your customers run?

 

I can understand that getting your hands on software to test can be expensive/hard!

Personally, I use SolidWorks and know for a fact that a quadro card doesn't produce visual glitches that are seen time to time on Geforce cards. From an endpoint view it doesn't matter but it's a little distraction that can add up over time. My other colleagues use Abaqus, which seems to run okay regardless but hates AMD cards under Linux (it's a cross-OS programme).

As for the MRI scanning software my colleague uses, it's proprietary! However, the micro-CT scanner we have has some software which can be used for a one month trial period, the name of which I've written down somewhere. However, you'd need data (TGA files) to construct a 3D model and typical model files consist of around 7GB of 2MB files.

 

Funny that this has come up. I've been posting in the hardware section, responding to a question about CAD GPU's.

In responding to that thread, I found this: http://www.opengl.org/discussion_boards/ubbthreads.php?ubb=showflat&Number=284014&page=1. Basically someone has created an OpenGL benchmarking tool.

Worth looking into, it might help in a comparison.

 

Thanks for the link Fingers66. Unfortunately, from what I can see the benchmark that's being talked about there is synthetic, not a real app, something we try to steer clear of on bit.

As, while x many more transfers per second or MB/sec may be a good thing, what really matters is how much faster do the apps that you run complete tasks.

Still, right now it's better than what we have, which doesn't ammount to much more than 3D renders (which are more CPU than GPU limited).

 

You are right, the "real world" benchmarks all use actual applications - which aren't cheap.

http://www.opengl.org/resources/benchmarks/

 

I know it's hardly ideal but would the openGL portion of Cinebench do the job?

 

Indeed, still I've got a few of them lying around the office for purposes like this. If anybody has any workloads/models they'd like to share for benchmarking purposes, I'll take another look at our editorial calendar and see if I can schedule some time to experiment with them.

Cinebench uses the same codebase as Cinema4D, which is used in a lot of films/games so it's actually a pretty good benchmark. I'd like some more tests though too.

 

I'd happily make some models for you to use but alas my own pc is borked, plus i'd need to know what versions of what software you had.

There was a site, whose name annoyingly eludes me that had scenes available to download for benchmark purposes [though i think that was for render times rather than viewport performance].

And i know Cinebench is based on 4D, that why i suggested it. I know you also use it for cpu benchmarking. 'Hardly ideal' was more referring to it only being one test.

 

Great, hopefully that's something we can explore in the future

That's another issue we need to fix too; as its very easy to measure render times, but in most apps that's CPU limited not GPU limited, so is great for Xeon and Opteron reviews, but not much cop for Quadro and FireGL reviews.

Still, as a race we thive on challenges!

 

It looks like there may be models included in the SPEC benchmarks: http://www.spec.org/benchmarks.html.

 

^this, plus the H80

maybe do a labs test for closed loop water coolers to see which is best as an out of the box unit

 

Agree, I also want to see the Mugen reviewed. It's one of the top coolers on many other review sites, and is about half the price of the Frio.

 

Mine is short and sweet.

Razer Hydra please.

 

Corsair 400R

 

i would love a watercooling masterclass the mag has been dry about this subject...?