MandelbrotJr

Negligible at best. Pretty much the only times you are using your HD while working in Max is when opening the scene file and when your renderer is accessing any texture maps. Sure the overall computing experience will be more snappy, although not even close to what you would get while using an SSD. The only time I'd consider using RAID over a SSD is while serving texture maps to several clients in an small to medium render farm. Of course there are tons of possible bottlenecks when dealing with these deployments, but you get the picture. Overall, if you're looking for raw speed nothing will come even close to an SSD from an storage standpoint and it's getting cheaper, specially if you're thinking in real HW RAID.

For all of you that are researching to buy a new machine, Intel just announced the new Xeon 5600 line of processors, including 6 core variants. Here you have the list http://www.intel.com/p/en_US/products/server/processor/xeon5000/specifications IMHO these are a much better buy than the i7 980X from a price/performance perspective. For the "money is no object" crowd you can buy a pair of X5680, with 12 cores/24 threads at 3.33 GHz it will probably shatter quite a few rendering records.

The Core i7-980X is a 130W TDP chip, there is no way to fit a chip with that thermal profile in a mainstream notebook. We'll need to wait until Gulftown tricles down to a more manageable 50sh. For comparison the fastest mobile chip right now is the i7-920XM and it's a 55W TDP unit.

Well, they are both good boards but while the Intel s5000xvn has the 5000x chipset the Supermicro X7DWA-N has the 5400. There are a couple of differences (ie. PCIE 2.0, etc) but the biggest one is that the 5400 can support CPUs with the FSB up to 1600 Mhz while the 5000x can only go up to 1333 Mhz. So CPUs with 1600 FSB like E5462, E5472 can only work on the suppermicro mb. Here are the links http://www.intel.com/Products/Workstation/Motherboards/S5000XVN/S5000XVN-overview.htm http://www.supermicro.com/products/motherboard/Xeon1333/5400/X7DWA-N.cfm http://www.intel.com/products/server/chipsets/5000x/5000x-overview.htm http://www.intel.com/Products/Server/Chipsets/5400/5400-overview.htm

Great Info Andrew, I specially like the way you summarize tons of tech details into a couple of key variables for the non-techie crowd. You might have already thought about this, and its certainly no easy task, but it would be great if you could get to benchmark your builds (HW loan maybe) with some of the most used sw. Tom's, Anand and a few others publish workstation hw reviews from time to time but there is no single site out there that specializes in this (specially unbiased ones). You could even create a few "Performance Indexes" to normalize the performance of the builds and make it much easier for people to grasp the performance gains vs price. Keep up the good work

You're right, I meant regular Radeon cards. FireGL/Pros are very very good cards, in fact they usually outperform Quadros in Max by a long shot, but my experience with Radeons and OpenGL is not a happy one, most probably driver related. Now, just to be clear Photoshop is NOT using CUDA to accelerate the viewports, just plain old OpenGL via AIF. http://kb2.adobe.com/cps/404/kb404898.html There is a Photoshop plugin sample shipped with CUDA 2.0 but that is a long shot from being an end customer thing, mostly geared towards devs. There are quite a bit of scientific, CFD, financial apps, encoders, etc built on CUDA, but for pure Arch Vis I think the only ones using it are VRay RT, Octane & Furry Ball. I might be missing something but it's a short list anyway. Overall, CUDA days are numbered, OpenCL & DirectCompute are gaining fast and I'm not sure anyone would be starting new development on CUDA now that even Nvidia is backing OpenCL. We'll see...

Just wondering, why are you guys only mentioning NVIDIA cards ? I mean if you're using OpenGL in any way shape or form I would stay far away from any ATI cards, but since Max has lost it's OpenGL roots a while ago in favor for DX I'd think ATI might have some serious advantages here. I mean for about 70usd more than a GTX 260 Core 216 you could buy a 5850 that will literally run circles around pretty much any Nvidia card in DX. The only other thing that I might consider is if I have anything dependent on CUDA, otherwise the AMD field seems to be much greener these days. Just a thought

BTW if you need to retrieve your W2K key, here you go, it's free http://magicaljellybean.com/keyfinder/

Tom, you shouldn't need to spend much money given the HW you already have. If you and your buddy need to work on the same files and implement some kind of pipeline it might make sense to convert one of your nodes into a file server, the cpu and ram usage of a file server is minimal so it won't degrade the rendering performance much. Now the issue is that XP only accepts 10 concurrent inbound connections so you can't server more than 9 machines. The good news is that W7 has a 20 inbound connection limit so it can serve all of your network. You just need to setup some shares at one of the W7 nodes turning it into a file server and point everything to it. Of course you'll need to soup up the storage, add some redundancy and implement SOLID backup strategy (RAID can never be counted as backup). Here you have some tips on how to configure the OSX clients to access windows shares http://articles.techrepublic.com.com/5100-10878_11-6127492.html On your questions 1 Windows Server costs a fortune, is it necessary for networking more than 10 farm pcs, even if the server is a mac? Neither a windows server or ANY kind of server is need for pure networking, you can keep everything IP based and share the stuff as I mentioned earlier. Now if you want centralized user administration, policy based permissions, etc etc then you'll need a true server. 2 Can I use a PC server? I know nothing about mac and the other guy isnt too tech-savvy. Windows networking is based on Netbios (the basic stuff at least), nowadays every modern OS supports it out of the box. Again it depends if you just want to share files or build a full blown network. 3 Is a mac server better in this situation? If you really really need a server I'd opt for a windows one, specially if you're not tech savvy. 4 Is it going to create problems having both mac and PC on the network? Nope 5 I have an old dual P3 machine with a copy of windows server 2000 on it. It is operational and I guess I could just throw some better network card and HD's into the box in RAID....would that be satisfactory? More than enough HW for a file server. You might want to invest is a HW RAID card if you're planning to drive a lot of throughput. 6 Could I continue with my work-group, add one of my drives as a drive on my colleagues Yup, as I explained above OSX can natively access windows shares. Let me know if you have any doubts

The Dell M line are pure workstations, but if you just want rendering power and not the full ISV certification and quadro cards you can probably get by a Dell Alienware M17x. It's a gaming notebook, but for $ 2000 you can a Q9000 quad-core, 4Gb RAM, 17 inch screen and a GTX 260M that will run circles around pretty much any mobile quadro in DirectX. For $ 3100 you'll get a Q9100 (2.26Ghz), 6GB RAM & 1200p screen.

For those who wanted to know how this numbers are generated you might want to check this site : http://global.bsa.org/idcglobalstudy2007/ These guys are the ones who commissioned the study. For those who don't know, "The Business Software Alliance is the voice of the world’s commercial software industry and its hardware partners before governments and in the international marketplace", you get the picture I always find these threads interesting because I experience all of this from different perspectives. I live in Argentina so I would be considered something between 3rd and 2nd world and while I haven't really done serious work in this industry I've been around for a long time. At the same time I work for IBM as a WW Enterprise Architect. The "globalization issue" that is deeply impacting this industry is exactly the same vehicle that most global companies have been using for the past decades to increase their profit. Heck IBM is roughly paying me 12usd / hour (which is a good salary in Argentina since we're 4 to 1 to the dollar) while at the same time charging 100usd/ hour for my services, you do the math. Same thing for companies building stuff in china for cents per unit. The IT & Manufacturing world has been adjusting to this for decades and it's pretty stable by now, the rules are clear. I think the Archviz (a much newer industry in itself) has "just" started this process and it will take a while to settle. Will be a top notch render be charged 5k in the future ? Don't think so. Will a top notch render be charged 40usd in the future ? Don't think so, the market will eventually regulate the lower end. I know it's not completely accurate to compare the IT world to Archviz, specially since the entry barrier to provide quality service is wildly different in both industries, but I do think it's the natural process for any industry adjusting to the new environment. Interesting times, I think the real challenge won't be a matter of producing the best or cheapest renderings, I think the challenge will be providing the best VALUE to the business that need this services and find new and creative ways of providing similar services to new customers who don't know they need them yet.

Good Point, SW prices from big corporations (Adobe, Microsoft, Autodesk, etc) have more or less the same price than in the US. HW prices are higher but not that much, 15% plus aprox. You can get XSI 6 Foundation for around u$s 550. But you are correct, that's the biggest investment that you need to face when starting this kind of bussiness, the ROI from the HW/SW purchase is much lower. The good news is that I can affort to pay myself much less compared to someone living in the US/UK from the same pure income and devote a bigger share of the money to cover for the operating expenses and pay up the SW/HW. Let's say I earn u$s 2000 a month (just a figurative number), I can take 1000 for me and expend the other 1000 paying the SW/HW and the rest of the expenses. The end result is the same overall. There are very good goverment sponsored credit lines for small business starting up, and tax cuts for capital purchaces (HW/SW). Also my SO earns as much as I do, that certanly helps

Hi, first let me tell you that this is one of the most interesting threads that I've read in the forum. While I'm haven't posted much, I've been a regular reader of the forum for more than two years and I think it would be interesting for the community to read a bit about the "other side". It might appear that I'm getting off topic, but bear with me for a while... I'm 29 yo and I live in Buenos Aires, Argentina. While not technically overseas it's a low cost country by any standard. I've been doing Architectural Visualization for 6 years as a hobby, and it's something I love to do. The interesting thing about my position is that I work for IBM, yep the big blue. I've been working for 7 years in IBM Argentina, 6 of them as a Global Resource (aka supporting processes in other countries). Right now I'm the enterprise architect for a brand at a WW level, in charge of making sure that all the systems and processes keep working from a technical standpoint and acting as a project manager for several high profile efforts. My manager is in charge of the whole IT operation at a WW level for that brand. I travel to the US 7/8 times a year and I have very good friends over there. And while this might sound pretentious, we are VERY good at what we do compared to US/European or WW standards. I earn about u$s 900 a month, and that is a very decent salary in Argentina. Just you give you an idea of how much you can do with that kind of money, I can pay the rent for my 2 bedroom flat almost 4 months. I have peers (as in reporting to my manager) in the US earning 5/10 times that much. So while any amount of money can sound as "too low" or "too high", it all depends on the context. For a IBMer working in Bangalore u$s 900 is a FORTUNE. So, and to get back to the main topic, globalization is here and it's not going away, but it's not only guys working on their homes undercutting US/UK visualization companies, it's also Fortune 500 companies saving billions of dollars and enriching their workforce by using top talent in other countries. There is people who is taking advantage of this global trend to undercut and just aim for the quick buck, and also there is people who love what they do and aim for quality, reliability and teamwork. This is the same everywhere. I'm been thinking for a while about opening my own company an earn a living working in visualization. And trust me, never ever I though "hey I can charge 50 bucks for each image, undercut some UK guy who's working his ass off to match my quality/price balance and force my way into the global market". Will my prices be lower that UK guy/company ? probably, due to the fact the cost of creating those images is lower (not the HW/SW, but education, taxes, salaries, expenses, etc are). And trust me, it's not going to be easy for me too, as I have a TON of competition in my own country that will try to undercut me. I really think that in this global market, both PRICE and QUALITY OF SERVICE are the two big drivers so I will need to get my "product/service/price" as a whole to be better than the rest if I am to succeed. It doesn't really matter where you live, the recipe for success is the same. Well, that's it. I hope these lines will give you a brief glimpse of what's in the mind of someone who just happened to be born in another country but shares the passion for a really great profession.

The clear benefit of the "renderfarm" approach is being cost effective. For example, the biggest baddest x86 machine money can buy right now is the Boxx APEXX 8 priced at about u$s 80.000 for the top of the line config (16 Core, 128GB RAM, 1.3TB HD, etc). With that amount of money you can buy a $5k more than decent primary workstation, and lets say 50 1.5K C2D nodes with enough RAM. The later approach will give you at least 30/40 times the performance of the single machine for the same price. Of course this is an extreme example, but the point is the same, a lot of cheap machines is ALMOST ALWAYS faster than a single monster machine. There is the matter of test renders, but a 3/4k machine will usually be more than enough for test renders. just my 2 cents

Well, lets not forget that we're not talking about HUGE increase in resolution. In the computer world we've been using resolutions equal or greater than 1024x768 for years, 720p is not that much of a leap from that. I often forget that in the TV world you still have a LOT of people looking at 486 lines of vertical resolution. Yes it will mean an increase in rendering times, but lets also remember that in the past year or so dual-core procs have reached a mainstream price, and that trend is going to continue for the foreseeable future with more cores on the way. I think HD is going to be a really big thing for us in the future, and although we will need to keep on eye on the detail, we're much more suited to producing content in HD than other industries because our data source is intrinsically scalable.