Another long post from me, sorry y'all ...
Have you asked the vendor of your application what they recommend? Many vendors do benchmarks internally, and they may have some guidance.
Side note, you're likely going to be using registered memory for dual MP or Opteron.
Animator, just to be contrary to my colleges...
I assume you're unlikely to even notice differences of less than 10%, and they are probably not worth paying a price premium to achieve. Even 20% is arguable!
On the "real-world" (application simulation) 32 bit tests I have seen at anandtech, tomshardware, cadalyst, etc. I would expect Athlon 2800+ MP and Opteron 240 to be well within 20% of each other, probably within 10%. Depending on the application, the Athlon MP 2800+ may be faster.
anandtech.com 3ds Max R5 benchmark
Comparing single Opteron 244 (1.8GHz) vs Athlon XP 2200+ (1.8GHz non-Barton )
equal for 1 of 5 3DMAX tests
within 10% for 3 of 5 3DMAX tests
within 20% for 1 of 5 3DMAX tests
tomshardware.com 3D Studio Max 5.1
Single Opteron 244 (1.8GHz 2x166) = 124
Athlon XP 2400+ (2.0GHz 133/2x133) = 115
Less is better, so the 2.0GHz Athlon is approx. 7% faster in tests than the 1.8GHz Opteron, for an 11% clock speed advantage, pretty close.
An Athlon MP 2800+ runs at 2.16GHz, but does not have dual channel memory.
So, I suggest that *clock for clock*, Athlon XP is mostly within 10% of Opteron for these 32 bit application tests.
These are *not* conclusive as the Athlon is using dual channel memory which is not available for MP. But at the *same* clock speed single CPU's are close, and an MP 2800+ runs 50% faster (2.16GHz) than an Opteron 240 (1.4GHz).
I looked at the CADLYST reviews
' ">http://www.cadonline.com/revie...dware/1103workstation/, and they were also inconclusive too, but more relevant as they do benchmark MP 2800+ on 3ds MAX 5.1.
See the last page (Table 1: Workstation Features
' ">http://www.cadonline.com/revie...ation/1103features.pdf) for a break down of results.
The Opterons are 246 (2.0GHz), so I've tried scaling the CADALYST MAX3d 5.1 benchmark numbers, using only Opteron clock speed (1.4GHz/2.0GHz), the Athlon 2800+ MP, is within 11% of the two Opterons:
@Xi Opteron 'scaled' 240 = 38.26 (246=54.66 scaled 1.4GHz/2.0GHz)
Monarch G64 'scaled' 240 = 35.69 (246=50.99 scaled 1.4GHz/2.0GHz)
Hypersonic Athlon MP 2800+ = 34.62
This suggests that the extra instructions supported by the Opteron are not critical (noticable) to this application.
I realise this is misleading as the processor clock isn't the only factor, but this is only an estimate.
So, conclusion from benchmarks - very close, and so inconclusive!
Based on these results, Athlon MP 2800+ is unlikely to be worse than 11% behind a 240, and on some tests, may even be faster, for your application. I assume other factors close to equal if you are buying the machines at the same time.
What is interesting is the significant improvement of the dual Opteron 246 in the CADALYST tests over the single CPU tests at anantech and tomshardware, this may be other parts of the system, like the graphics card.
Whatever you do, be careful. The CADALYST tests compared two dual Opteron 246 systems, apparently with the *same motherboard*, graphics card, and similar memory. They got more than 10% difference in performance (which surprised me!
I haven't investigated this, but this scale of difference is as significant as the Athlon MP 2800+ vs Opteron 240 difference.
Also, many of the dual Opteron motherboards are $200-$300 more than MP. I can see applications get *much* more performance from $200-$300 of extra RAM (or disk, or graphics card) rather than slightly faster CPUs. Your rendering problem *may* be an example of this.
If you were a software developer building AMD64 software, had 64 bit software, or just had to have bragging rights, buy Opteron 240; you have no choice.
But, I think *today*, $ for $ Athlon MP is better value at the low cost end. This does show that the AMD marketing guys are getting their pricing about right;-)
You may want to consider selling parts of either system in 8 months when, I hope, Opteron is running DDR400, or better, and low cost Opteron is clearly faster than MP.
Can you scrape together the extra cash for 242 or 244 and avoid the stalemate?-)