IMAC i5 vs i7, Need your advice

It sounds like you'll get more use from the better graphics than from the better CPU... go with the i5.

The only thing is that the 5750 doesn't support double precision at all whereas the 4850 does. People have tested both with OpenCL samples and found the 5750 to be faster but if 3D computation in some apps moves to the GPU, some algorithms will require double precision support. The i7 iMac also scores 9800 vs 7700 with the i5 on Geekbench so the i7 is nearly 30% faster CPU-wise. If you have a 15 minute render on the i5, it will take about 12 minutes on the i7 - if you do animations, it could be say 4 hours vs 3 hours. The i7 has hyper-threading so it behaves like an 8-core machine but the i5 has no hyper-threading. This means 3D render buckets can start concurrently and max out your RAM (you should get 8GB btw).

Ideally you'd just get a lower-end Mac with an overclocked headless PC that you build yourself for the best value:

http://www.youtube.com/watch?v=ZOBsawCJut0

1GB of VRAM is good on the i5 iMac and it will run cooler but post-production 3D rendering won't be done on the GPU for a while yet and will likely need double-precision support, although some apps are moving there so you're going to be dependent on the CPU for getting renders finished.

In terms of gaming performance, they are about the same so 3D viewport speed will be the same. The VRAM just means more textures and geometry can fit in memory but very few things will max out 512MB.

Personally I'd go for the i7 but there's not a whole lot between them. I reckon you'll want to upgrade at the end of next year anyway as Ivy Bridge will allow them to move to 6-8 physical cores in the iMac with 12-16 threads.

I suppose the i7 variant you are looking at is an older model.

The i5 runs cooler, which is a bonus if you don't have air conditioning in the summer. 3D modeling is not really CPU-bound anymore. Animation, sure, but you seem to be more on the engineering side than the art side -- otherwise, you'd have to be stupid to use AutoCAD. Even for engineering work, it is a questionable decision to use it.

Anyway, CPUs these days are more that enough for engineering modeling. If you are using subdivision solids, for one thing it is imprecise and art-centric, and for a second thing it is a memory-bound task more so than a CPU task. If you are a rockstar and are using splines directly , then you don't really need much power or memory at all. The same goes for basic solids and surfaces modeling, which are typical of engineering CAD. Ray-tracing is a linear process, so hyperthreading is truly pointless for photoreal rendering. Getting a kick-ass renderer like ElectricImage will do a lot more for render speed than will a marginal increase in CPU performance.

There may be something to say about using the card that supports doubles in HW. Graphics are rarely double-precision (in fact, a lot of graphics are "half" precision), but sometimes scientific stuff uses double-precision. It's probably a non-factor, but you should look into benchmarks for the 5750 vs 4850.

That's what people who use single-threaded rendering engines like EIAS Camera say . Cinebench alone counters that statement for both multi-core and HT:

http://vr-zone.com/articles/does-cor...lps-/6160.html

You would use subframe distributed rendering with Renderama to get a similar effect on a single machine.


I would agree in principle that faster software trumps faster hardware if the software is proven to be significantly faster. Trouble is, with something as complex as 3D software, you can never really get empirical data to prove it's the case whereas you can with hardware.

Benchmarks show you will absolutely get a certain increase by buying the i7 but you couldn't say that buying EI vs Mental Ray will absolutely give a speedup. In any case, both machines are the same price so even if the engine is faster, buying the i7 + EI would surely be the best option.

The GPU is really the deciding factor here IMO. The 5000-series will draw less power, support newer standards like DX11, has more VRAM but the 4850 has DP support. I think the i5 might have an extra USB port and you might get an older mouse with the old iMac. It's the age-old problem of buying an older higher-end or a newer lower-end machine.

The two machines in question perform very closely so I don't think it makes a huge difference. The MacWorld benchmarks for each place them very close together:

http://www.macworld.com/reviews/prod..._i5266ghz.html
http://www.macworld.com/article/1531...mac_tests.html

Raw CPU tests show just a 15-20% advantage to the i7, although their tests are usually too short to be useful. The GPUs are on par. The i7 has VT-d support whereas the i5 doesn't - this is for directed IO in virtual machines so direct GPU support running AutoCAD in Parallels for example. But these technologies are in early stages:

http://forum.parallels.com/showthread.php?t=96488

The iMac won't be a long-term purchase when you need it for 3D so it really doesn't matter which one you get. Given that the i5 will have more newer hardware architecture vs the i7 having higher-end features you aren't likely to use, the i5 is probably the more sensible choice.

Sandy Bridge is just 4 weeks away so if you aren't desperate for a new machine ASAP, it may be worth holding back. Usually people base upgrades on the release cycle averages but the cycles can vary a lot - 5-11 months. The time between the the last update and January will be 5-6 months.

Apple can update the MB, Mini, MBP and iMac at once near January. Then focus on the iPad in March, then the iPhone in June, then the Ivy Bridge updates in October/November including the Mac Pro.

Is cinebench ray tracing? It doesn't look like it. For non-raytracing, I don't doubt that HT can offer some bonus. HT is just a trick that allows the cache not to be dumped as often between context switches, so if it is done right, a meaty single thread will actually run slightly better than two smaller threads. It's just that most threads in SW aren't very big, so HT makes sense.

This is pretty good advice, although I dunno if Apple will get the quantities it needs to update to Sandy Bridge right away. I don't see January even if Intel is claiming a faster ramp. How long was the lag between mobile nehalem and the iMac update? About a month and a half?

I'm thinking March is the likely timeframe. Who knows, the iPad 2 might come earlier anyway.

Still, assuming your PC isn't decrepit you can probably wait long enough even if the price is likely to change a little.

If you want, get the iMac 27" with HD5750 1GB AND custom order a 2.9Ghz Core i7 to go with that. It is only $200 USD more. You are spending that much on a system, go for the full spec.

Yeah the Nehalem launched Sep 8th '09 and the iMac came out October 20th '09 but there was a big design change moving to the 27" and quad-core.

The 2010 update had more minor design changes, the chips arrived July 18th and the iMac was updated July 27th.

I don't think they have any reason to even make minor design changes to the iMac. They should just be able to drop the new chips right in.