Months have gone swiftly past since the initial announcement that SPECapc for 3ds Max 9 would be replaced by SPECapc for 3ds Max 2011. The previous version of SPECapc for 3ds Max 9 certainly did show how good their own professional graphics cards performed (on specific systems shown within the SPEC website). Until now the website was dominated by the Tier 1 vendors. However, things change. The new SPECapc for 3ds Max 2011 benchmark removes these limitations. New, revisited SPECapc benchmark provides so much information not only for the System Integrators and VARS but to the actual end-users as well.

Many organizations tend to use their own in-house view sets to test each of the professional graphic cards supplied out by both AMD and NVIDIA. With systems becoming more complicated, the SPECapc committee sat back with Autodesk to produce a unified benchmark that not only tests the professional graphics card but the complete system I/O as well. Therefore, whenever and wherever possible they reduce the possible bottlenecks and this in turn allows end-users to meet the demands of their customers on time and within budget with the best possible system for each task requirement. The benchmark and submissions are not just aimed at the Tier 1 companies, but to every System Integrator (SI) and Value Added Reseller. This is turn means that when supplying out to companies and end-users utilizing the benchmark allows all to finely tune their systems for Autodesk 3ds Max.

Autodesk 3ds Max and 3ds Max Design applications provide powerful, integrated 3D modeling, animation, rendering, and compositing tools that enable artists and designers to more quickly ramp up for production. The two versions share core technology and features, but offer differentiated experiences and specialized toolsets for game developers, visual effects artists, and graphic designers on the one hand, and architects, designers, engineers, and visualization specialists on the other.

SPECapc for 3ds Max 2011 Home Menu
SPECapc for 3ds Max 2011 Home Menu

Therefore we see one powerful package of software for those who do not fully understand what it consists of. With a package of this magnitude it in turn requires a benchmark to fully encompass the primary system I/O components. The new SPECapc for 3ds Max 2011 sets out to complete this accomplishment.

For those new to SPECapc/gpc and the varying committees and what they do we have to give a brief insight to you; so each understands what takes place behind these secretive closed door sessions.

Standard Performance Evaluation Committee
Many questions are asked who or what consists of the SPEC committee and its sub-division SPEC/GPWG (SPEC Graphics and Workstation Performance Group), as this is a non-profit organization that sponsors the development of standardized, application-based benchmarks that have value to the vendor, research and user communities. Supported by the main project group members. Therefore for those not in the know, it consists of AMD, Dell, Fujitsu, Hewlett-Packard, Intel, Lenovo, NEC and NVIDIA. Simply put; these Tier 1 vendors sit down around the table regularly and thrash out what we annually look forward to – the new professional benchmarks (apc/gpc) that stretch professional graphic workstations to the limits.

It allows the professional end user the chance to evaluate their own system in situ – or possible new system about to be purchased an exceptional idea of guidance on how it will perform to their own individual needs. It also means that with the returned figures shown within the SPEC website from its members users can see which system or professional graphics card suits their needs. Therefore when it?s time to make that all important purchase from the Tier 1 or some of the more famous system integrators guidance can be gained on how it might perform. One important point that some System Integrators and VAR?s overlook is that they can submit their individual systems for consideration and their benchmark results reviewed and, if accepted, posted upon the SPEC web site. Once this achievement has been made the individuals results are seen many Corporate and Government CTOs and IT managers who rely on SPEC results as a key component in their decision-making process for purchasing new computer equipment. The payback can be very considerable indeed.

SPECapc Brief History
The SPECapc (SPEC Application Performance Characterization) group was formed in 1997 to address performance evaluation based on popular workstation applications. Current members include AMD, Dell, Fujitsu, Hewlett-Packard, Intel, Lenovo, NEC and NVIDIA.

Since its inception, SPECapc has worked with independent software vendors (ISVs) to ensure that its benchmarks represent what application users experience in the real world. The credibility of SPECapc benchmarks is proven by the number of SPECapc results reported by publications, websites, vendors and analyst organizations throughout the world. In addition to SPECapc for 3ds Max 2011, SPECapc provides benchmarks for Autodesk Maya, Lightwave, Pro/ENGINEER, SolidWorks and Siemens NX.

SPECapc for 3ds Max 2011 - Building flyby
SPECapc for 3ds Max 2011 – Building flyby

Therefore we can see that the committee have been around for considerable amounts of time and the current serving committee members have been within the industry for many years with a wealth of experience.
Changes between SPECapc for 3ds Max 9 and 3ds Max 2011
Moving on and the big questions that are now being asked – what is the difference between SPECapc for 3ds Max 2011 benchmark and previous SPECapc for 3ds Max 9?

The new features and tests that have been added and major upgrades have been made in SPECapc for 3ds Max 2011, include:

  • Updates that exercise new functionality implemented by Autodesk in 3ds Max 2011.
  • An improved user interface that makes it much easier to configure and run tests.
  • New and revised test cases for CPU, GPU, modeling and interactive graphics.
  • Two separate versions of the benchmark, one for professional and one for personal use.
  • Increased level of testing for shading and rendering, including in the professional version taking advantage of the Autodesk Quicksilver engine that uses both the CPU and GPU for accelerated rendering.
  • In the professional version, a city scene model containing nearly 32-million polygons that tests extremely large model creation and playback.
  • Automated benchmark results compilation for the professional version.
  • A pay-for-download system that collects payments online.

One of questions asked is what are the differences between the professional and personal versions of the benchmark. The professional version of SPECapc for 3ds Max 2011 sells for $495 (USD) and is designed for those serious about better performance. This includes users for whom performance increases can yield significant bottom-line value, consultants who make their living helping clients seek out maximum performance per dollar spent, and vendors that rely on performance measurement as a means of improving products, boosting sales, creating competitive advantage, and enhancing relationships with customers.

SPECapc for 3ds Max 2011 Professional version is a comprehensive test suite that exercises diverse aspects of 3ds Max 2011 performance, including modeling, interactive graphics, CPU and GPU. It includes 58 separate tests – including tests that build and render a city scene containing nearly 32 million polygons (faces) in real time – providing a complete performance picture for systems running 3ds Max 2011. Results from the professional version can be published publicly and submitted to SPEC for publication on its website.

SPECapc for 3ds Max? 2011 Personal version sells for $20 (USD) and provides a basic, easy-to-use benchmark that generates a single score. It runs a subset of the tests found in the professional version, testing performance based on smaller models. This is ideal for someone who is seeking more information about 3ds Max 2011 performance, but who does not have the need for a comprehensive test suite using large models and does not require publication of test results.

SPECapc benchmarks have always been free though benchmarks such as SPECapc for 3ds Max 2011 are extremely expensive to create, test and validate. As it currently stands – industry companies have always paid the bills for SPECapc benchmark development, but in recent years industry consolidation has reduced membership revenue, forcing SPECapc to investigate new revenue models. With SPECapc for 3ds Max 2011, SPECapc is asking customers who benefit from its benchmarks to help contribute to the cost of their development. This is a realistic professional approach now and reduces the danger of one-sided approach we have witnessed with the other standards body. For all the current multimedia benchmarks online whether professional or advanced, payment is met to meet the development costs of new builds and for a majority of end-users this is perceived as not an issue. After all everybody wants to know how fast their systems are, therefore this step taken now by the SPECapc Committee should this be no different.

How SPECapc for 3ds Max 2012 Works

SPECapc for 3ds Max 2012 Professional Version Menu
SPECapc for 3ds Max 2012 Professional Version Menu

The new features in SPECapc for 3ds Max? 2011 include:

  • Updated tests based on new functionality in 3ds Max ?2011
  • An improved user interface that makes it easier to configure and run tests
  • Increased level of testing for shading and rendering in the Professional Version, including use of the Autodesk Quicksilver engine for accelerated CPU and GPU rendering
  • Automated benchmark results compilation in the Professional Version

A single score is reported for the Personal Version. Results for the Professional Version are derived by taking the total number of seconds to run each test and normalizing it based on a reference machine, in this instance a Dell Precision 690 workstation with 2.0GHz Intel Xeon 5130 processor, 4 x 4GB FB-DIMM DDR2 SDRAM (ECC) memory, NVIDIA Quadro FX 570 graphics card, and 80GB Seagate 7200RPM hard drive. The normalization process ensures a scoring system where a bigger score is better. Composite scores for the Professional Version are reported for CPU, GPU and large-model (city scene) performance.

SPECapc for 3ds Max 2012 Brutal

The recommended memory is 16GB for the Professional Version and 8GB for the Personal Version. The Professional Version supports only systems running the Microsoft Windows Win7 64-bit operating system. The Personal Version is unsupported, but it is recommended for Win7 32-bit and 64-bit.

SPECapc thanks Autodesk for contributing content and expertise to this benchmark, and to independent animators who created models, including Mike O?Rourke of Fritz Studio, Andy Murdock of Lots of Robots, Gary M. Davis of visualZ, and Zack Baker.

System Setup and Software Used
Our array of "in-house" test systems has now been fully completed as over the forthcoming months, many exciting things will be happening here. We worked with Supermicro and Intel to set up excellent parts which in turn means we can now demonstrate the best possible results for upcoming computer components. This in turn will demonstrate to you the end user just what to fully expect when you purchase systems of the like we have here in situ.

With the supplied parts on show today these main parts now form part of our "mainstay in house" test systems supplied courteously by AMD, Akasa, Intel, Crucial Memory, NVIDIA and Supermicro; which could be found in most high-end studios and enthusiast workstation scenarios. These are very reliable and sound workstation platforms which have an abundance of scope for upgrading, which we have voiced on many occasion.

For SPECapc for 3ds Max 2011, after much deliberation we decided to move forward with our ultimate workstation builds: Supermicro’s X8DAI Rev2 motherboard armed with a pair of the superb Intel Westmere 3.33GHz CPUs. Thrown into the mix as these new boards are just hitting the streets, we put into place Supermicro’s new X9SCA board that support both Intel’s Sandy Bridge Xeon E3-1280 (3.5GHz) processors for workstations encompassing the Intel® C204 PCH chipset. Interestingly enough both surpassed themselves providing exemplarily results. At times we could not tell the difference when they were stacked up side by side.

From the professional graphics cards front, taking into account the actual brief given we took to the fore the most up to date professional graphics cards from each company listed below.

Test System – Single Processor

  • Intel Xeon E3-1280, 3.5GHz, 8MB Shared Cache, 5GB/s QPI
  • Supermicro X9SCA
  • Akasa Nero 2 1155 HSF
  • 16GB (4x4GB) Crucial DDR3-1333MHz Unbuffered ECC
  • 250GB Intel 510 Series Solid State Drive (6GB/s)

Test System – Dual Processor

  • Supermicro X8DAi Rev2
  • 2 X 3.33GHz Intel Xeon X5680 Nehalem-EP, 12MB Shared Cache, 6.4GB/s QPI
  • 2 X Noctua NH-U9DX
  • 24GB (12x2GB) Crucial DDR3-1333MHz Unbuffered ECC
  • 300GB Intel 320 Series Solid State Drive (3GB/s)

These formulized builds used will mean that the results shown within will be extremely clear to the reader.

On a side note and very important one. Many self-build corporate and studio users maintain the philosophy to utilize one motherboard which provides them with an effective cost performance solution. Time is money as too are extra peripherals needed to supplement a motherboard. Once the motherboard has reached end of life in the usage, motherboards of this nature then find them stripped down and placed into a 2U chassis as a dedicated server or render node. Therefore the life span of these motherboards and supporting processors is extremely good, which brings us back once more to exceptional value for money. 

From the professional graphics cards front, taking into account the actual brief given we took to the fore the most up to date professional graphics cards from each company listed below.

Tested Graphics Cards

  • AMD FirePro V8800 2GB GDDR5
  • AMD FirePro V9800 4GB GDDR5
  • NVIDIA Quadro 4000 2GB GDDR5
  • NVIDIA Quadro 5000 2.5GB GDDR5
  • NVIDIA Quadro 6000 6GB GDDR5

Overall, six graphics cards feature 18.5GB of ultra-fast GDDR5 memory, or an  average of 3.1GB per card. All cards used the same memory standard – GDDR5 which de facto is the standard for high performing graphics processors.

Benchmarks and Software Used 64-bit Mode

  • Microsoft Windows 7 Professional 64-bit
  • Autodesk 3ds Max 2011 64-bit
  • SPECapc for 3ds Max 2011 64-bit Professional Edition

Each set of tests has been applied on the clean system hard drives shown above to ensure that no residue drivers were left installed with all updates/patches applied. A test/render has been completed many times over different periods of the system uptime. Maintaining the fair play rules of SPEC Hyper-Threading and Turbo Boost have been enabled, with the memory being left in its default status of Auto. Tests have been conducted in accordance with the resolutions stipulated within the test run rules of 1920×1080 @ 59Hz / 60Hz in 32-bit color. Results that have been shown within this article are from the application/benchmarks first run in accordance with the SPECapc fair play rules.

To maintain completely unbiased attitude we have given a complete dedicated page to each companies professional cards throughout the test procedure and therefore remain neutral by not stacking up the cards side by side.

AMD Professional Graphic Cards on Dual Socket Intel Xeon System

FirePro V7900 performs greatly
FirePro V7900 performs greatly

AMD FirePro V8800 2GB - 17.53 GPU Score
AMD FirePro V8800 2GB – 17.53 GPU Score, 26.26 Composite

AMD FirePro V9800 4GB - 17.05 GPU Score, 26.21 Composite - You can see the effects of double-stacked memory (higher latency)
AMD FirePro V9800 4GB – 17.05 GPU Score, 26.21 Composite – You can see the effects of double-stacked memory (higher latency), as the card lags (but barely) behind FirePro V8800 – which is identical card other than the amount of video memory

Cayman architecture efficiency at work: Even though it has less processing cores, Cayman-based V7900 outperforms older, high-end Cypress-based brothers.
Cayman architecture efficiency at work: Even though it has less processing cores, Cayman-based V7900 outperforms older, high-end Cypress-based brothers.

Test in detail: All boards are leveled besides the Large Model GPU, where V7900 leads by 45%
Test in detail: All boards are leveled besides the Large Model GPU, where V7900 leads by 45%

First out the stable door was to be that of the FirePro V7900, it’s only been a few weeks since we looked at it and the initial impressions had been good. Today is no different as it romped home within the Large Model GPU test taking everyone by complete surprise.

The new GPU features internal architectural improvements (VLIW4 instead of VLIW5 concept) has made substantial differences against the rest of the pack. The actual Large Model Composite result was a substantial outcome therefore it shows that the FirePro V7900 has got the steam to deliver. We must point out that this driver release is a new beta from AMD.
NVIDIA Professional Graphic Cards on Dual Socket Intel Xeon System
NVIDIA Quadro 4000 2GB: 24.45 on GPU composite and 26.64 on Large Model leaps ahead of AMD boards sans V7900 in Large Model Composite
NVIDIA Quadro 4000 2GB: 24.45 on GPU composite and 26.64 on Large Model leaps ahead of AMD boards sans V7900 in Large Model Composite

NVIDIA Quadro 5000 2.5GB brings in GPU composite score to 28.40, while Large Model achieves excellent 32.37
NVIDIA Quadro 5000 2.5GB brings in GPU composite score to 28.40, while Large Model achieves excellent 32.37

NVIDIA Quadro 6000 6GB takes no prisoners: 29.45 in GPU Composite and 38.56 in Large model, 20% higher than any other card on the test
NVIDIA Quadro 6000 6GB takes no prisoners: 29.45 in GPU Composite and 38.56 in Large model, 20% higher than any other card on the test

Composite Scores for Quadro boards on Dual Xeon
Composite Scores for Quadro boards on Dual Xeon

NVIDIA Quadro Combined Composite results
Complete scores for Quadro boards on Intel Dual Xeon setup. You can easily see where the 6GB frame buffer took the lead

Not to disappoint, all NVIDIA Quadros came flying out of the stable door with the full emphasis on the need for speed. What is interesting is on how good the scaling is within each card as we go further up. The bigger the card, the bigger the increase in scaling - within all areas. The raw power of the Quadro 6000 shining through on the Large Model GPU test. This will appease many end-users thinking on taking the plunge in this area.
AMD Professional Graphic Cards on Single Socket Intel Xeon System

FirePro V7900 on a single Xeon sytem: 20.92 and 35.84 scores raise suspicion that SPECapc for 3ds Max 2011 is CPU-limited
FirePro V7900 on a single Xeon sytem: 20.92 and 35.84 scores raise suspicion that SPECapc for 3ds Max 2011 is CPU-limited

FirePro V8800 also scores higher results than on Dual Xeon system
FirePro V8800 also scores higher results than on Dual Xeon system

FirePro V9800 results still vary between higher amount of memory and higher latency
FirePro V9800 results still vary between higher amount of memory and higher latency

Here is where things got interesting indeed, there has been much expectation on how well the new Supermicro X9SCA with the Intel® C204 PCH chipset and E-Series Xeon would perform. Impeccably. What we found interesting on our findings was that the actual FirePro professional graphic cards performed better on the single socket than the dual socket – as a clear evidence how Sandy Bridge architecture is more efficient that the older Westmere. Once more and still maintaining the lead over its older siblings in many areas was to be that of the FirePro V7900. Bear in mind this new card is AMD?s first of possibly many to come.
NVIDIA Professional Graphic Cards on Single Socket Intel Xeon System

Quadro 4000 on a single-socket system yielded quite high results: 24.04 and clean 30.00 in Large Model
Quadro 4000 on a single-socket system yielded quite high results: 24.04 and clean 30.00 in Large Model

Quadro 5000 continues to score really good and the lower CPU composite score doesn't affect the GPU ones
Quadro 5000 continues to score really good and the lower CPU composite score doesn’t affect the GPU ones

King of SPECapc for 3ds Max 2011 Hill: Quadro 6000 6GB GDDR5
King of SPECapc for 3ds Max 2011 Hill: Quadro 6000 6GB GDDR5

Composite results for NVIDIA Quadro cards: note almost perfect scaling
Composite results for NVIDIA Quadro cards: note almost perfect scaling

Complete scores for NVIDIA Quadro cards on SPECapc for 3ds Max 2011
Complete scores for NVIDIA Quadro cards on SPECapc for 3ds Max 2011 – no doubt who is the overall winner here

Once more the sheer brute strength of the Quadro came to the fore maintaining superb scores on Supermicro?s new X9SCA and the E3-1280 Xeon processor. The cards scaled all the way through each specific test just as one would expect. Surprisingly enough the GPU rendering and GPU shaders on this platform were somewhat faster than upon the Dual Socket Xeon. This is a good keynote for the end-users that are considering moving over to Sandy Bridge. Needless to say, we’re eagerly expecting to see what level of performance dual-socket Sandy Bridge E processors will bring to the market.
Conclusion
Just like SPECviewperf 11 last year, we’re seeing that all the new benchmarks from SPEC are becoming more user friendly but still ensures that final outcome is extremely accurate. The new SPECapc for 3ds Max 2011 is a superb upgrade from the previous generation of SPECapc for 3ds Max 2009. Its execution is more smooth, with the data sets loading much quicker and the final result breakdown far more easier to all too readily understand. We firstly must turn to those unsung heroes that have spent many grueling months putting this all together, without doubt kudos to you all.

The SPECapc committees have given a benchmark that completely stretches not only the graphics card but the actual whole system I/O. For those out in the wild who complain that nothing is ever given back, please think again as these guys and girls who make up the SPEC/GWPG and its project groups – SPECapc and SPECgpc; give up much of their free time to attend these committees and oversee the professional benchmarks so many take for granted. There will be a few that will complain of the cost but as we stipulated within the introduction this is a realistic approach now taken by the SPECapc committee. To reiterate, everyone is very comfortable paying 100?s of dollars for professional benchmarks developed by BAPCo and Futuremark Corporation etc, then why should this be no different… This is a realistic professional approach now, and all walks of industry that supply hardware systems utilizing 3ds Max should be supporting the benchmark.

For the more adventurous who wish to see the full remit, try sitting watching a complete professional run that can take up to 3-4 hours professional card dependant of course. It’s only after this you will truly understand just how much grueling and demanding effort has gone into this build of the new SPECapc for 3ds Max 2011. Below are a few test render shots grabbed whilst the test is underway.

SPECapc for 3ds Max 2011 Renders  

It brings to the fore the question many of the more larger System Integrators, even the smaller System Integrators companies and Value Added Resellers must now be asking "why should we submit a submission, after all, the Tier 1′s dominate these sorts of things". It?s just not so the case. Not just a year ago; several companies completed and successfully met the criteria and their rewards where great. It did involve a little bit of extra work behind the scenes, though they did all reap the reward of very successful orders. Many System Integrators have different build plans which are claimed to be fast and reliable, it just takes that courageous judgment to step up to the starters gate and that leap of faith that your product can succeed. It can be done successfully. The systems we have demonstrated within today are what many can easily produce to run 3ds Max. At the end of the day, it can be achieved successfully.

The actual results we have shown today have been most interesting indeed. Starting with the CPU’s, without a doubt in certain aspects the Dual Xeon still currently maintains its crown within the actual CPU Rendering and CPU Computing sectors. However the new C204 PCH chipset and E-Series Xeon have made significant impacts as the results within the Large Model CPU test has shown remarkable leaps forward, a full three point gain is a pretty significant leap. There are exciting times ahead with the speculated new Dual Xeon sockets and of course the advent of the new PCI Express 3.0 standard.

Within the professional graphic cards results we saw some amazing sets of results. We have been on about results, therefore below is a screen shot of the combined scores of a completed run.

All scores combined: SPECapc for 3ds Max 2011

The young pretender, AMD FirePro V7900 really set the pace for all the cards from AMD and it did show some very interesting results indeed over its older siblings. Its score within the Large Model GPU result and Composite test on both test system platforms took everyone by complete surprise. It’s amazing what can be achieved in a year within this industry and how rapidly technology moves from these companies. AMD’s driver is still within the beta stages and should be completed soon and therefore – we will revisit in due course.

NVIDIA Quadro cards made a formidable noise which was made right across the board with some very exciting numbers to be seen. Each of the Quadro cards scaled as was to be expected within a test environment. It is without a doubt the Quadro 6000 showed it?s true might in the correct light of day. The Large Model GPU test scores are phenomenally astonishing to say the least and, puts to rest the skeptics and pessimists who claimed that the card was too big with too much memory. The driver team at NVIDIA have been working around the clock to get these optimizations correct and kudos to you all.

The year ahead, it’s going to be very interesting indeed. All those wonderful new hardware parts that are speculated upon the horizon will make differences within certain viewsets and render outputs. Of course each new driver and hardware release from both AMD and NVIDIA will bring with it more advancement in the final viewset and CPU results. Therefore we all have to watch these spaces very closely as the going is about to get very tough and exciting indeed.

Ultimately the new SPECapc for 3ds Max 2011 completely stretches the full system I/O. With careful planning of production systems in conjunction with the new benchmark will be pretty spectacular meeting the demands of this ISV software. Those with the fastest and most stable platforms will end up winning outright those lucrative orders.

BSN* Value Award 2011: AMD FirePro V7900 2GB GDDR5
BSN* Value Award 2011: AMD FirePro V7900 2GB GDDR5


BSN* Editors’ Choice 2011: NVIDIA Quadro 6000 6GB GDDR5
BSN* Editor's Choice 2011: NVIDIA Quadro 6000 6GB GDDR5

BSN* Editors’ Choice 2011: SPECapc for 3ds Max 2011
BSN* Editor's Choice 2011: SPECapc for 3ds Max 2011