The latest from the DEVELOP3D Blog:
Published 26 February 2009
Posted by Greg Corke
Despite the emergence of multiple CPU cores in individual workstations (the standard is now 4 cores) most CAE applications still can’t harness all the available power, probably never will until there’s a major architectural change, and many max out at two cores. The way round this is to use clusters, or to put it simply a series of desktop computers directly connected together with some clever stuff.
Blue Ridge Numericsis the latest CFD vendor to promote its own cluster solution, which works with Microsoft Windows HPC networks. (I’m not hugely familiar with cluster technology, but I do know that Microsoft introduced its cluster technology a few years back, in response to the success of Linux in this sector.)
Blue Ridge’s CFdesign high performance computing (HPC) Module works directly with Windows Server 2008 HPC Edition, and with a little tinkering can turn two or more desktop workstations into a mini cluster. As with a single workstation, there is a limit to the performance gains you can get but Blue Ridge is quoting a maximum reduction in simulation time of 550% for a four node, 16 core setup. From what I can ascertain it looks like there might be a little exaggeration here as the baseline appears to be taken from a single core workstation.
Blue Ridge told me last year that in a single multi core workstation one can expect speed increases in CFdesign of up to 140% (Intel) to 160% (AMD), though I would imagine Intel has now got faster with the introduction of the Core i7, which receives its data directly from system RAM rather than going via a front side bus (FSB).
So by my rough calculations, the performance increase from a single multi core workstation to a four node, 16 core cluster, would be about 350%, which still is not to be sniffed at (I’m sure Blue Ridge will correct me if I’m wrong here).
Percentages aside it’s certainly a very interesting technology and I look forward to test-driving it in the near future. I’d better dust off my copy of Windows HPC cluster for Dummies first though.
View comments (4 comments)
Published 24 February 2009
Posted by Greg Corke
The big news at the tail end of last year was the launch of Intel’s brand new Core i7 chip. Codenamed Nehalem, the Quad Core chip features a brand new architecture, which represents one, if not the biggest architectural shifts in Intel processors for ten years.
Core i7 systems are shipping now from the specialist systems builders such as CAD2, Xworks and Scan, and we should start to see Core i7-based workstations from the likes of HP, Dell, Fujitsu Siemens and Lenovo in the next month or so.
So what is all the fuss about? There are three major architectural changes to
Core i7 that not only look good on paper, but should have a real impact on the way users work with CAD/CAM/CAE and rendering applications, so let’s have a look at each of these in turn.
At the heart of this new architecture is a change in the way the chip accesses memory. Instead of the CPU communicating with the memory via the Front Side Bus, Core i7 can receive data directly from the system RAM. If this sounds familiar, that’s because it is, as AMD pioneered this integrated controller strategy with its Athlon & Opteron processors a few years back.
With Intel’s Front Side Bus architecture, which is used on the Core 2 Duo and many generations before, there was a lot more latency when accessing memory. Now with Core i7, applications that access a lot of memory, frequently, will see a benefit. This is why AMD’s Opteron has remained a popular choice with certain CAE users, despite it being slower in most mainstream applications.
The other change in the memory architecture is that the new memory controller has three channels to the RAM which means that Core i7 systems will work best when memory modules are in multiples of three, as opposed to two. This means we are likely to see workstations with 3GB, 6GB and 12GB memory instead of the usual 2GB, 4GB, and 8GB.
All Core i7 CPUs have four cores as standard, but they also feature a technology called HyperThreading, which simulates additional threads so each chip actually has eight logical cores.
HyperThreading first came to market with the Pentium 4, but was abandoned for the Core 2. It uses spare CPU cycles on each physical core to simulate additional cores, and these can be seen when you bring up the Windows System Performance Dialogue.
The technology only works with certain multi-threaded applications, and can cause confusion when a process assigns itself to a ‘logical core’ even when there is a physical core sitting around doing nothing. Our limited tests show that it does make a small but significant difference in rendering applications such as 3ds Max.
TURBO MODE / OVERCLOCKING
Core i7 features a new Turbo Mode technology which can automatically adjust the speed of the cores dynamically. The chip can literally switch off those cores that are not being used and channel additional power to the remaining cores.
Intel claims that for single threaded applications (of which most CAD applications are) the speed of a single core can be boosted by around 400MHz.
While Turbo Mode can dynamically adjust the speed of the CPU, specialist workstations manufacturers are looking to get more out of each piece of silicon by overclocking or permanently increasing the speed of chips.
With Core i7 and indeed Core2 Duo, Intel has built in a lot of headroom into its chips. Some say this is because it has no real incentive to sell faster CPUs at this moment in time, because it could jeopardise future sales if the performance leap is too high.
The good news is that those in the know are able to get more out of the chips for no additional cost, safely overclocking them by around 20%.
Overclocking has never really been used in the CAD/CAM/CAE sector, simply because reliability has been deemed more important than performance. However, even with standard cooling solutions, specialist workstation vendors are now offering overclocked systems. But this is not pony tailed geeks in bedrooms with soldering irons, the system builders are extremely confident that the silicon will not be damaged by overheating and this is being backed up with three year warranties.
At DEVELOP3D, we don’t expect overclocking to be embraced wholeheartedly by customers, simply for fear of unreliability. However, if confidence grows, and specialist workstation manufacturers continue to push overclocked systems, it will be very interesting to see what impact this has on the Tier One Vendors as the likes of HP, Dell and Lenovo will have to play by the rules and ship systems at Intel’s published speeds. And with the top-end 3.2GHz Core i7 chips costing around £700 it’s not only a performance advantage that we’re talking about here. Specialist system manufacturers are already offering 2.66GHz Core i7 chips clocked up to 3.2GHz for under £300, so price/performance could also become a major differentiator.
Core i7 is a huge leap forward for Intel, introducing a number of new technologies, which not only improve efficiency in multi-threaded applications but enhance performance in single threaded applications. As with any new technology, prices are high at the moment, but deals can be had on overclocked systems and expect significant cuts later this year.
View comments (2 comments)
Published 24 February 2009
Posted by Greg Corke
Back in 2005 the workstation sector was all about AMDand its Opteron CPU. At that time the Opteron wiped the floor with anything Intel could throw at it and all the major Tier One workstation vendors, with the exception of Dell, took on the powerful chip to head up their workstation lines. But then Intel introduced the Core 2 Duo and things changed overnight. AMD lost its leadership position and Intel began outgunning AMD across the board. Since then AMD-based workstations have only been seen in niche sectors, targeting price/performance sweet spots or users of specialist CAE applications.
This month, however, AMD introduced a brand new quad core chip called the Phenom II X4, which it hopes will take the fight back to Intel. Available in speeds from 2.5GHz to 3.0GHz, the Phenom II X4 is built around AMD’s Direct Connect architecture, which means it receives data directly from the system RAM, rather than going via a slower Front Side Bus. This is a technology that AMD originally pioneered with its Opteron processor and one that Intel has only just ‘borrowed’ for its Core i7.
To put this new technology to the test we managed to get hold of an extremely rare 2.8GHz Phenom II X4 925 courtesy of specialist workstation builder XWorks. Nestled alongside an AMD ATI FirePro V3750, 4GB RAM, a 250GB hard drive and Windows XP SP3, the compact workstation is designed to offer excellent price/performance and at £745, the price is certainly competitive.
While the cost of an AMD Phenom II X4 925 chip is comparative to a 2.67GHz Intel Core i7 920, the big saving comes through the ASUS M3A78-CM mainboard, which with a retail cost of around £50, is around a quarter of the price of most Core i7 motherboards, which are particularly expensive at this moment in time.
But what about the performance? In terms of CPU-specific tests we benchmarked with 3ds Max Design 2009, an application which makes full use of multiple cores when rendering scenes. The performance was almost 50% slower than the overclocked 3.2GHz Core i7-based Scan 3XS workstation we tested last month, but that doesn’t tell the whole story as far as CPU performance is concerned. 3ds Max is one of the few 3D applications to offer finely tuned support for HyperThreading, a unique feature of the Core i7 which boosts multithreaded performance by simulating additional cores (click here).
The most interesting results, for CAD users at least, came when testing graphics under SolidWorks and Inventor. To begin with we didn’t quite believe the SolidWorks 2009 score of 17 Frames Per Second (FPS), simply because it’s the fastest we’ve yet to see in any machine, but repeated tests verified our initial findings. The Inventor score of 3.3 FPS was more in line with what we’d expect, slower than the Core i7, but impressive all the same.
From our tests it’s hard to pass a definitive judgment on the Phenom II X4, simply because it doesn’t support HyperThreading, which gives Intel’s Core i7 an advantage in our 3ds Max benchmark that it wouldn’t have in other applications. However, for CAD applications our 3D graphics tests, which are very much linked directly to CPU performance, indicate that it’s certainly in the same ball park as the Intel Core i7.
While AMD looks to have an interesting technology on its hands, particularly when compared to Intel’s Core i7 in terms of price/performance, one mustn’t forget Intel’s Core 2 Duo, which is now being sold at extremely competitive prices, even at high GHz. And if dual core is not enough because you have specific multithreaded rendering or analysis requirements, then Intel’s Core 2 Quad has also come down in price since the introduction of Core i7.
The long and short of it all is that AMD has delivered a chip, which while not setting the workstation sector on fire, certainly warrants closer attention. While it is unlikely that Tier One vendors will back the Phenom II X4 wholeheartedly, the chip is likely to be picked up by agile system builders like XWorks meaning at long last there’s an AMD chip that should be considered alongside Intel when choosing a workstation, particularly when the purse strings are tight.
View comments (1 comment)
Published 18 February 2009
Posted by Al Dean
As I type this, the world’s media (probably all 12 of them) are sat waiting for Autodesk to unveil its product releases for the next year. As with these things, Autodesk love to group everything together and make a Huge Splash – a good idea, but you often find that while there’s a concerted effort to announce the products together, the actual shipping of code lags somewhat. Either way, whether you think that’s right or not, let’s look at the highlights:
#1: Inventor 2010: Autodesk is integrating the plastic parts design tools it’s been working on for the last few years – and it looks like the mould design tools they’ve also been beta testing out in China for at least year are finally going to ship in product, rather than on Labs.
Other news is that simulation has had a boost. Recently acquisitions see Autodesk’s armoury expanded greatly and this is the first wave of new simulation tools. There’s also some bundling going on with Inventor LT (the free version of Inventor for part design), along with AutoCAD LT.
STL output in Alias has always been flakey unless you used a workaround with spider – this fixes that.
#2: In a right world, this would be Number 1: Autodesk is making the Alias line of products available on the mac OS X platform. THIS, is HUGE news. There’s also been work done on packaging (it’s got new bundle and naming), new STL output, new rendering and materials tools and a lot more control over surfaces.
#3: Showcase is getting the ray-tracing technology acquired from Opticore some time ago. There’s also greater support for workflows with both other Autodesk products (such as Max an Maya) and some pretty slick looking design variant comparison tools. it also has a matched material library with Alias. Oh and its going to $995 instead of five grand.
#4: Autodesk Moldflow: there’s a simplified version of Moldflow on its way. One thing I have to take issue with is the idea that Moldflow can help take someone that’s “not an expert in plastic injection moulding right through to manufacture.” Ok. That’s really not a good idea. Moldflow is good, but it can be dangerous in the wrong hands (by dangerous, I mean expensive).
NB: there should be more pics, but they weren’t. Of anything except Alias and Inventor. A Shame.
View comments (8 comments)