Home > Articles by: JP Dormehl

Author Archive

Where we came from

athlon-xp1700+My first PC ran on an AMD AthlonXP 1700+. It had a Palomino core running at 1.43Ghz, a 266MHz front side bus, and 256KB of  L2 cache. But that wasn’t what started my infatuation with technology. That was due to my friend’s Pentium 1 133mhz and Voodoo graphics card. Back then, playing StarCraft, Theme Hospital, and The Settlers 2 til the early hours of the morning was a common occurrence.

While a thorough history of CPUs is beyond the scope of this article, let’s take a look at a few titbits of the industry that brought us to where we are today.

CPUs have come a long way since the 2300 transistor, 740kHz, 4-bit Intel 4004 was released in 1971. At the time, AMD was just a maker of logic chips.

It wasn’t until 8086 that Intel introduced the x86 architecture that we all love (and sometimes hate). It was the 8088 that was used in the IBM PC though, running at 4.77MHz with 16KB to 256KB of RAM. Then, on the 12th of August, 1981, The IBM PC  5150 was released. From its use  of (mostly) off-the-shelf components, an open architecture, and the release of the IBM PC Technical Reference Manual, many cheap clones soon began to enter the market. The 5150 did not come standard with a HDD, but prospective buyers had the option of either a  floppy disk or cassette system.

In 1991, Intel then released the first 32-bit x86 processor, the Intel 386DX, which AMD soon reverse-engineered and named the Am386.  The Am386 reached higher clock speeds than Intel’s 386, at 40MHz and 33MHz respectively. Rather startlingly, a 40MHz Am386 armed with a 40MHz 387 Math co-processor would outperform even Intel’s new and expensive 486 platform.

pentium

Pentium architecture

Next, we have the renowned Pentium processor. Its 3.1 million transistors ran at a blisteringly-fast 60MHz on a 0.8µm process, and boasted a 50MHz FSB. The Pentium had a few improvements over the 486, including the ability to complete more than one instruction per clock cycle. However, the processor suffered from a bug in its floating point unit that caused errors while performing certain floating point division calculations. Perhaps ironically, the  media furore surrounding this bug, combined with Intel’s “Intel Inside” marketing campaign, soon made Intel a household name.

Fast-forward to 1999: AMD becomes the first CPU manufacturer to break the 1GHz barrier with their K7 Athlon. With the Athlon 64, AMD let the aging front side bus fall to the wayside in favour of Hypertransport. The Athlon 64 also brought with it 64-bit extensions for the x86 instruction set (called x86-64) which allowed full backwards compatibility with existing 32-bit applications. This implementation differed radically from Intel’s 64-bit implementation (IA-64) in the Itanium, which had poor x86 32-bit performance. Eventually, Intel incorporated AMD’s instructions under the name EM64T.

In 2000, Intel released the Pentium 4 which used the infamous RAMBUS RAM, which they later switched for SDR, then DDR. The Netburst architecture was supposed to scale to 10GHz, but I suppose Intel decided that having multiple suns in our solar system was a bad idea.

That’s where we’ll stop for now. In the next article, we’ll have a look at the processors currently on the market.

Essential (and free) software

We all have our special few programs that we cannot live without and always install after every time we format our PCs. Below are a few of the applications that I hold dear, and all are free to download.

Firefox (7.6MB)

When looking at the market share for October 2009 [as per w3schools.com - statistics vary significantly according to the polled user demographic -Ed.], IE8, IE7,IE6, Firefox, Chrome, Safari, and Opera are at 12.8%, 14.1%,10.6%, 47.5%, 8.0%, 3.8%, and 2.3% respectively, Firefox is currently the top Web browser. With the plethora of plug-ins available for the browser, it’s not hard to see why.

OpenOffice.org (150MB)

Since April 30th 2002, when Open Office(OO) 1.0 was released, the product has taken great strides and is a competent alternative to the Microsoft Office suite of products. OO consists of multiple components, with Base, Calc, Draw, Impress, Math, and Writer being the OO versions of Microsoft Access, Microsoft Excel, Microsoft Visio, Microsoft Power Point, Microsoft Equation Editor, and Microsoft Word, respectively.

Thunderbird (6.4MB)

Thunderbird is a Mozilla-developed e-mail client for a complete replacement of Microsoft Outlook.

SuperAntiSpyware

Super Anti-Spyware (7.2MB)

I have a lot of appreciation for this application. While paid-for versions of SAS are available, the free version does everything you need it to.

Foxit PDF Reader (5.11MB)

Who likes waiting an age for PDF files to open? Not me. Foxit Reader is a lightweight, lightning-fast PDF reader. Foxit

Reader also has support for tabs, so you can have multiple PDF files open in one instance of the reader, which makes life much easier when working with multiple PDF documents.

K-Lite Codec Pack (Size Varies)

K-lite is perfect for someone (like me) who plays a lot of media on their PC and doesn’t want to worry about which codecs they have installed. It also comes with Media Player Classic, so you can play all your videos in one player instead of requiring multiple players for multiple formats. The K-Lite pack comes in multiple versions (Basic 4.7MB; Standard 10.4MB; Full14.2MB; Mega 23.2MB) with the Mega Pack having support for Real Media files. For a complete run-down of each package’s capabilities, please look here.

VLC Media Player (17.2MB)

As good as Media Player Classic is, I just prefer VLC. Whether it’s because I just know all the shortcuts or because it just works, I’m not sure. VLC also can do a lot more than just play media files; for a full feature list take a gander here.

imgburn1

ImgBurn (1.48MB)

For anyone who is tired of Nero’s bloat, ImgBurn comes to the rescue.

Pidgin (13.8MB)

The default installer has support for 16 different IMs, and more with the use of plug-ins. You can use Pidgin to chat on MXit from your PC and look a lot busier at work than if you were using your phone. A slightly older version of Pidgin with the MXit plug-in can be found here.

7-Zip (1MB)

A tiny application with support for about a gazillion formats, and better compression ratios than WinZip.

Another notable mentions include Teracopy, Daemon Tools Lite, AVAST, DVD Decrypter, Free Download Manager, PC Tools Firewall Plus, Guru3D Driver Sweeper,and nlite.

SSD isn’t everything

GSkillFalcon1A hard disk drive is the slowest component in your PC. While SSDs are freakishly expense (the G.Skill Falcon SSD 64GB costs around R2600) they do vastly improve the performance of the PC equation. But is price the only reason people aren’t more interested in them, or do people just not care about HDD performance? I did a huge poll (among 5 of my friends) and found that most people would always choose extra capacity over extra performance.

Well, what can you do if you want extra performance but aren’t willing to shell out the R2600 for a G.Skill Falcon? What you can do, is get yourself a Western Digital Black drive. I recently bought the 750gig version, which costs around R200 more than a 7200.12 1TB from Seagate. Or, for a little less than the price of the 1.5TB Seagate drive, you could buy a 1TB WD Black. Is it worth it losing 500 gigs of space for better performance? I’d say so. A faster hard drive will make your operating system boot up and shut down faster as well as making your games load faster.

Once you go black, you don't go back.

Once you go black, you don't go back.

Anandtech did a nice preview of the Seagate Barracuda XT 2TB: SATA 6Gb/s versus the WD 2TB Black drive. In four out of the six tests performed, the WD black drive was faster than the Seagate. The WD drive was between 3% and 28% faster, while in the two tests that the Seagate was faster, it was only between 3% and 7% faster.

Another way to get better HDD performance is enabling AHCI (Advanced Host Controller Interface) mode in the BIOS, but bear in mind that only Vista and Windows 7 have native support for AHCI mode. For Windows XP, you will have to slipstream the AHCI driver in. AHCI mode enables features such as hot-plugging and native command queuing (NCQ), which allows your hard drive to rearrange the requests for best performance. Much like if you and two friends had to get on an elevator and push the floors 3,12 and 6, it would be terribly inefficient for the elevator to travel to floor 3 then 12 then 6. NCQ allows the drive to rearrange write/read commands that are transmitted randomly in order to optimize the movement of the drive heads. So going back to the elevator example, the requests will be rearranged so that the elevator goes to floor 3 then 6 then 12. On average, AHCI mode improves performance by about 5%.

Interconnection costs

We all hate the fact that cellular phone calls are so expensive, but have you ever given any serious thought as to why? We have some of the most expensive call costs in the world mostly due to high interconnect rates. The interconnect rate is what telecommunication companies charge each other when a call from an external network is terminated on their own network.

mtn-logo-smallIn the early 1990s, before Cell C entered the market, Vodacom and MTN had an interconnect rate of around 20c per minute. This was judged to be fair, and was based on a cellular market a lot smaller than the one we have now. Yet when Cell C entered the market, the peak interconnect rate was raised to R1.25 a minute. That’s an increase of over 500%. Why would the interconnect rate need to be increased? With growing economies of scale, it really should not have.

The answer is simple. A high interconnect rate severely disadvantages smaller players in the market.  When you are a smaller player, the majority of calls placed will be off-net calls and with a high interconnect rate, there is very little in the way of wiggle room with regards to prices; dropping the call rates below the interconnect rate just isn’t feasible. Dropping on-net rates will have a negligible effect on attracting clients, as a small user base does not make cheap on-net rates attractive.

Research ICT Africa Director Alison Gillwald suggested an interconnect rate of R 0.25, instead of the current peak-time cellular interconnect rate of R 1.25 per minute. He said this was a conservative figure which leaves a fair margin for profit.

vodacom_logoVodacom generated R8.63 billion  through interconnection in South Africa for the financial year ending 31 March 2009, up from R7.94 billion a year ago.  This is the company’s second largest revenue stream after “airtime and access”, and is far higher than the R5.97 billion generated through data services, or the R 5.19 billion from equipment sales. For the financial year which ended December 2008, MTN generated R6.95 billion in interconnect revenue, up from R6.34 billion for the previous twelve months.  Similarly to Vodacom, it is also the second largest revenue generator for MTN, after airtime and subscription revenue. MTN’s interconnect revenue is more than its data and SMS revenue (R3.59 billion) and cellular telephone and accessories sales (R3.12 billion) added together.*

ICASA and the Parliamentary Portfolio Committee on Communications have been making a lot of noise lately regarding the interconnect rate. Whether anything actually gets done is another matter, as ICASA is generally seen as a toothless regulator. One also has to question why the government is now so interested in dropping the interconnect rate after government’s cash cow, Telkom, no longer has shares in its cash cow – Vodacom.

The next few months in the telecommunications industry will definitely be interesting to watch.

*Interconnection figures via MyBroadband

Research ICT Africa Director Alison Gillwald suggested an interconnect rate of R 0.25 instead of the current peak-time cellular interconnect rate of R 1.25 per minute which he said was a conservative figure which leaves a fair margin for profit.

Where to from here, nVidia?

microsoft-tegra-zune-hd-pmp

As most of you should know, nVidia doesn’t just manufacture video cards. It also has a motherboard chipset division, and produces mobile chipsets like Tegra, which is found in the Zune HD.  While ATI has been taking all the headlines with RV870 and DirectX11, and most of the nVidia partners are set to discontinue their GT200 cards (or already have), what will nVidia have to show to keep its brand in the mind of consumers?

Its ION chipset is definitely one of them.  ION is by far the best Atom chipset: with an onboard 9400, it can handle HD video and accelerate any program that supports CUDA. At the typical resolution paired with netbooks, the 9400 can even provide for low-detail gaming. At first, Intel tried to resist attempts by nVidia to get its ION chipset into the mainstream, preferring rather to bundle Intel’s old and power-hungry 945GSE Express chipset, but as of now, the HP Mini 311, Lenovo IdeaPad S12, Samsung N510, Acer AspireRevo, and Asus eeeBox EB1012 all make use of the ION chipset.

Tegra is a little marvel; it consumes less then 100mw at idle, and includes an integrated audio processor, a graphics processor and two ARM cores. Tegra’s application processor features an ARM11 processor core with a GeForce GPU, and dedicated image and HD video processing engines.  The integrated GeForce GPU features programmable pixel shaders and support for OpenGL ES 2.0 and D3D Mobile.

samsung-tegra-phoneIf one looks  towards Tesla, however, things get a bit shaky. NVidia is putting a lot of effort into something that only made them $10M last quarter.  However, the fields that nVidia will target, which include seismic processing, supercomputing, universities, defence and financial research, could end up bringing in almost $1B of business for the company. The seismic processing industry, for example, could use 32 Tesla S1070s to replace 2000 x86 servers, which could generate a saving of over $7M and 1100kW. As I touched on in my last article, Fermi is meant to be a GPGPU powerhouse.  The additional support for ECC, enabled C++, much improved double-precision performance and easier Visual Studio integration, should help nVidia market Tesla to a new audience, and the fact that Fermi is capable of addressing up to 1TB of graphics memory should be another boon for memory hungry applications.

As for their chipset business, it’s not as easy to quantify. Intel isn’t allowing nVidia to produce DMI chipsets, which puts the company at a huge disadvantage in that segment, and producing hardware for virtually every Apple Mac released today is an advantage for Intel. NVidia’s AMD-based  motherboards generally have more features at a cheaper price, but more and more companies are choosing to go with ATI chipsets for their AMD-based machines. NVidia is slowly being squeezed out of the desktop market, it seems. Even ION isn’t safe: Intel’s next generation of Atom CPUs will have integrated graphics, which could see nVidia pushed out of there as well.

GT300 is Fermi

It looks like we’ll getting a repeat of the 4-series versus GT200 battle, with a twist. Where nVidia went with a huge, power hungry core and the performance crown, ATI went for affordability and power efficiency on the smaller 55nm process, while nVidia were still using 65nm. ATI’s approach gained them a lot of market share, and now nVidia is on the back foot again. ATI were the first to get to the 40nm process with the 4770, and now they are the first with DirectX11 hardware as well.

nvidia-physxNVidia’s DirectX10.1 cards are coming in the form of GT220 and G210 on October 12 and the  GT240 in November. The GT220 will have 48 processor cores, a 128-bit memory interface and will be available in both DDR3 and DDR2. Reference clocks will be 615MHz for the core, 1335MHz for shaders and 790MHz for 1GB of DDR3 memory. As for the G210, it will have 16 processor cores and a castrated 64-bit memory interface. Reference clocks will be 589MHz for the core, 1404MHz for shaders, and 800MHz for DDR2 memory. With performance below ATI’s 4670, GT220 and G210 will have to be cheap to have a place in the market.

Nvidia hopes to change its fortunes with Fermi. Not going solely for gaming performance, Fermi will be heavily concentrated towards parallel GPGPU computing. With support for C++ and ECC memory, Fermi is definitely getting closer to CPU territory than other GPUs before. Therein lies the twist: nVidia is pushing CUDA and PhysX hard, and one cannot deny the improvements PhysX makes to Batman Arkham Asylum.

nvidia-gt300While on the topic of PhysX, those that were hoping to use an ATI card like the 5850 as the main display card and something like a 9600GT for PhysX should be fuming at nVidia right now. NVidia have recently disabled PhysX support in their drivers  if a card from another vendor is detected on your system. Quite frankly, it’s disgusting and will most likely backfire. People don’t like being forced to use what companies tell them to use – 3DFX learnt that lesson the hard way.

Getting back to Fermi, the chip will support DDR5 and should be bigger, hotter, and more powerful than the 5870. Even though the chip is based on a 40nm process, it will only be slightly smaller than GT200. The top Fermi card being released looks to  have 512 shader cores in 16 clusters, with 3.1Billion transistors, and will include DirectX11 support. The performance should be stellar – according to nVidia, this is the biggest architectural change since the G80 (8-series).

3 Fermi cards are set to get released at the end of this year. Most likely there won’t be any real quantities available, while ATI should have volume up within the next few weeks, as 3 months is a long time to get yields up. Notebook Fermi cards are only expected some time in 2010.

nVidia isn’t too bad off when it comes to DirectX 11 games, as DiRT 2, Alien vs Predator, and STALKER: Call of Pripyat are all being released late 2009 or early 2010.

ATI 5-series preview

We are on the cusp of a new generation of computer graphics. Directx11 will be released with Windows 7 at the end of October, and the software will be here this year as well. DiRT 2 for PC has been delayed so that Directx11 support can be incorporated into the game.

Today, we will have the release of AMD’s 40nm DirectX11 graphics cards, namely the 5870 and 5850. Just for some perspective, the 4870 and 4850 are both 55nm cards with 800 stream processors running at 750Mhz and 625Mhz respectively. With regards to memory, the 4870 runs DDR5 at 3200Mhz and the 4850 DDR3 at 1986Mhz.

ATrHD5870_angle2_lo_res

The 5870 will have 1600 stream processors running at 825Mhz with DDR5 memory running at 5200Mhz effective providing over 150GB/s of memory bandwidth. The 5850 will have 1440 stream processors running at 725Mhz with DDR5 memory running at 4000Mhz effective providing 136BG/s of memory bandwidth. Both cards will have 32 ROPs. The 5870 and 5850 will have a maximum power draw of  180W and 170W under load respectively and 27W when idle. The 5870 will come in 1GB and 2GB variants, while the 5850 will be released with only in 1GB form initially.

Just by looking at the specs, the 5870 should give more performance than the 4870×2. A 5870×2 is also planned, but apparently the drivers still need some polishing, and will be released about a month after the 5870. Power consumption might also be a problem. Even on 40nm, a 5870×2 card will have a power consumption of around 376W, which is 90W more than the 4870×2 on a 55nm process.

5-series cards for notebooks will be available around November, and that’s one place where 40nm will really come in handy. In a notebook, space, cooling, and power consumption all come at a premium. 40nm cards will be smaller, run cooler, and use less power, prolonging battery life.

ATrHD5870_angle1_lo_resWith the release of the next generation AMD graphics cards, AMD will be launching what they call Eyefinity. Imagine playing DiRT 2 across 6 monitors at a resolution of 7680 x 3200, and still having playable framerates. It’s possible. ATI has an agreement with Samsung to produce Eyefinity monitors which are extremely thin.

Things are not looking so well for nVidia on the GT300 front. Even if everything goes according to plan, GT300 is only expected in Q1 2010. Things haven’t been going well, though: rumours have been going around that yields are somewhere around 2%, which, quite frankly, is pathetic. A re-spin will add weeks to the release of the cards. ATI were smart to test out the 40nm process first using the 4770. They took what they learned and now have almost perfect execution in the 5-series. Obviously, we’ll have to see how availability is, first, but everything is looking good so far.

A look at SSDs, part 2

According to JEDEC specifications, MLC NAND flash can perform 10 000 writes before it starts to fail; SLC can perform around 100 000 writes. There’s no need to worry, however. If you have a 256GB MLC solid state drive, and copied around 7 gigs of data to the drive every day, it would take around 360 000 days – or 986 years – before each page has been written to 10 000 times, thanks to some clever wear levelling algorithms that get incorporated into SSDs. Each page on the SSD will get written to first, before the first page gets erased and re-written.

As explained previously, an SSD will get slower due to the pages filling with valid and invalid data. It is possible for a person to claim back performance by performing a secure erase using a tool like HDD ERASE – it will free every page on the drive giving you back your performance. And if you don’t feel like losing all your data, you can use Indilinx’s Wiper Tool.

OCZ’s Solid 2 uses 34nm Intel flash with an Indilinx controller.

OCZ’s Solid 2 uses 34nm Intel flash with an Indilinx controller.

A TRIM is as good as a holiday

What TRIM does is take the delete process one step further. Instead of the delete command ending with the OS, the command goes through to the SSD and the pointers to the data get deleted. So while the data is actually still there, you’ve told your SSD that you no longer care about it. The tool queries the OS for available LBAs, then commands your SSD to TRIM those LBAs, giving you a drive that’s virtually “as good as new”.  At the moment Windows 7 is the only Microsoft OS that supports TRIM. After Windows 7 launches (22/10/09), most SSD manufacturers intend to release TRIM-supporting firmware for their drives.

Indilinx is actually the first to officially support the ATA8-ACS2 TRIM command. Most new Indilinx-based SSDs have a 4KB random write speed of around 13.1MB/s; without TRIM support the 4KB random write speed drops to around 6.93. Once no longer new, with TRIM, you get around 12.9MB/s. For comparison, a conventional hard disk will sport speeds of around 0.3MB/s – 0.7MB/s.

OCZ’s Agility uses an Indilinx controller with either Intel 50nm flash or 40nm Toshiba flash.

OCZ’s Agility uses an Indilinx controller with either Intel 50nm flash or 40nm Toshiba flash.

JMicron controllers suck…

…and it’s all thanks to small file write latencies. The JMicron JMF602 controller suffers from slight stuttering. Your PC will pause for anywhere up to a second, then carry on as per normal. Even with JMicron’s updated JMF602B (with double the onboard cache) stuttering is still prominent. Average latencies are in the 500ms range – that’s around 100 times slower than your average desktop drive. Some manufacturers have put two JMF602B controllers in RAID configuration, to help bring average write latencies down to around 300ms, which is still far higher than Intel’s and Indilinx’s Barefoot controllers that have a 4KB write latency of 0.22ms and 0.34ms, respectively.

While not as bad, Samsung’s RBB controller is also lagging behind the cheaper Indilinx controller with regards to performance. While new 4KB random write performance may be 3x that of Western Digital’s VelociRaptor range, once each page has been written to, its performance drops even more. It’s just not worth it, especially when you take into account that the firmware isn’t user upgradeable.

A look at SSDs, part 1

Did you know that the main performance bottleneck in your PC is your hard drive? The hard drive is the one component that hasn’t seen any fundamental changes since it was first introduced in 1956 by IBM, and is long overdue for a replacement.

Enter the Solid State Disk (SSD). Providing sub-1ms access times, it brings performance that no standard hard drive ever could. SSDs are completely silent, require far less electricity to operate and, as they have no moving parts, are more reliable. An SSD will likely be the most meaningful PC component upgrade you will ever make.

Let’s first take a look at what makes a Hard Disk Drive slow. An HDD has 1-5 platters with a read and write head per platter. These platters are spinning discs with a thin layer of ferromagnetic material, usually 10-20nm in thickness. The write head magnetizes the material directionally to represent a “1″ or a “0″. The platters typically spin at between 5400rpm and 15000rpm, 5400rpm being common for laptop hard drives and 7200rpm being common for desktop hard drives. They have a random write speed of about 1.2 MB/s, and a sequential write speed of around 80MB/s – 100MB/s. Sequential read speeds are around 80MB/s – 100 MB/s when the drive is new, and at 10000rpm random access times are around 7ms.

hard drive platter

The fastest SSDs have an access time of around 0.11ms, and random write speeds going as high as 31.7 MB/s. Sequential write speeds on some models are as high as 240MB/s and read speeds are faster as well at around 260 MB/s, close to the limit of the SATA II specification.

SSDs do lose performance over time as well, due to the fundamental way in which NAND flash works. 1 or 2 bits of data are stored per cell (1bit for SLC and 2bits for MLC), cells are grouped into pages, being the smallest readable/writeable structure in a SSD, with 4KB pages being common today. Pages are grouped together into blocks, with usually 128 pages per block. A block is the smallest structure that can be erased in a NAND-flash device,  so while you can read from and write to a page, you can only erase a block. This is where the problem lies.

TorqXFor example, if you have a block with 384KB of data and delete 128KB, as with with hard disk drives, the data doesn’t get deleted, the OS marks it as invalid. So, to the OS you have 256KB “free” space while in actual fact there’s only 128KB free in that block (256KB valid data and 128 invalid data). Now if you try to save 256KB of data to that block, the initial 256KB of valid data needs to be copied to memory(or local cache if the SSD supports it), the whole block erased, then the 512KB of data written to the block. Even while taking this delay into account, SSDs are still significantly faster than standard hard disks.

In part 2 we’ll look at the SSDs available in the South African market, how you can claim back lost performance, and why you shouldn’t purchase a SSD with a JMicron controller.

The dual core vs quad core duel

Whenever someone considers building a new PC, the question always comes up: dual or quad?

It’s well known that games aren’t the pinnacle of multi-threaded software, understandably so. Dual core processors out-sell their quad core counterparts many times over, not to mention that Windows isn’t that great at juggling multiple cores.

Looking at local prices, a PhenomII x2 550 (3.1Ghz dual core with 6MB L3 cache) goes for between R1200 and R1300, while the Phenom II x4 945 (3.0Ghz also with 6MB L3 cache) goes for between R2600 and R2700. The argument could be made that, when speaking about a gaming PC, the extra spent money on the quad could rather be spent on a better graphics card. That ~R1300 could make the difference between a 4850 and a 4890.

So, if you only use your PC for gaming, a 3.1Ghz dual with a Radeon 4890 or Nvidia GTX 260, will give you better frames per second (FPS) than a quad with a 4850 or a GTS 250. Who only plays games on their PCs though? To anyone who does a lot of video editing or 3D rendering, the quad would be far more useful. While you are busy with other tasks, your graphics card sits there twiddling its thumbs, it’s your CPU that gets the work done, crunching those ones and zeros.

AMD_Phenom_X3_logoWhat makes things more interesting are the triple core processors. A PhenomIIx3 920 (2.8Ghz tri-core with 6MB L3 cache) can be had for between R1700-R1800.

Let’s compare identically-clocked dual, triple, and quad core processors. When encoding a song with iTunes, they all perform identically. Converting a video using MainConcept Reference shows  the triple core as being 33% faster and the quad core 48% faster than the dual. In AutoDesk 3Ds MAX 2009, the quad completed rendering a 1920*1080 frame 46% faster, while the triple core was 23% faster. An AVG scan completes 42% quicker using a quad and 23% quicker using the triple. In Winrar, the quad performs 25% quicker and the triple  20% quicker.

When it comes to gaming, things are slightly different. In Crysis, the triple core performed 17.4% better, with the quad giving virtually identical results. In Left 4 Dead, the triple core performed 9.7% better, with the quad improving marginally. World in Conflict showed virtually no difference; it is only when an AVG scan is performed concurrently with the World in Conflict benchmark does the quad really shine. The triple performed 6.25% better, while the quad showed a massive 156% increase in the minimum FPS over the dual.

Corei7

So when it games to gaming, the quad really isn’t necessary, a triple core would be my recommendation. Although I haven’t listed the percentage improvements, GTA IV and Far Cry 2 are two games that make use of the extra core that a triple core provides. With DirectX 11 also bringing improved multi-threading support, a triple core is the perfect middle ground if you, like me, want the best bang for your buck.


Advertisement

Advertisement

Login / Search

Latest games

Latest opinions

Advertisement