Slashdot Mirror
Can Legacy Dual-Core CPUs Drive Modern Graphics Cards?
MojoKid writes "A few weeks back, we discussed whether a new GPU like the GeForce GTX 660 could breathe new life into an older quad-core gaming system built in mid 2008. The answer concluded was definitely yes — but many readers asked to reconsider the question, this time using a lower-end dual-core Core 2 Duo. The Core 2 Duo CPU chip used was a first-generation C2D part based on Intel's 65nm Conroe core. It's clocked at 3GHz with 4MB of L2 cache and has a 1333MHz FSB. The CPU was paired with 3GB of DDR2-1066 memory. The long and short of it is, you can upgrade the graphics card on a six year-old dual core machine and expect to see a noticeable improvement in game performance — significant gains in fact, up to 50 percent or more."
At least not reliably.
The issue is PCI express 1.0 and 1.1 performance on 2.0 cards and later. Geforces have been known to crash using an earlier slot technology or on lower end systems. Maybe that has changed since the 9600GTX, but I switched to ATI for this reason. Even many Radeons are only tested with later hardware and instability and other bottlenecks happen as many games as Windows swaps video ram to the system ram even when there is plenty of ram available.
http://saveie6.com/
To save you a few clicks, here's the key conclusion (and much better said than the summary from /.) :
Intel Core 2 Q6600 chips aren't available new these days, but Ebay has a ton of them, regularly priced between $50-$70. (...) Is a new CPU worth the price? I'd say yes --especially if you've currently got a dual-core CPU in the 2.2 - 2.6GHz range. The combined cost of a used Q6600 and a GeForce GTX 660 should still come in below $300 while delivering far better performance than any bottom-end desktop you might assemble for that price tag.
Quem a paca cara compra, paca cara pagará.
I await the obviously conclusive "Can a Pentium M / Sempron be revived by a dual GTX680" article...
Exactly my thoughts. 50% increase in performance? Not really impressive when you look at the graphics card charts out there. GTX 260 has far from 1/2 the performance of a GTX 660.
According to PassMark:
GTX 660: 4038
GTX 260: 1123
So with only a 50% increase in performance, I'd say it's a waste of money. The bottom line is that modern processors, chipsets, & RAM will make a massive difference in performance for modern high end graphics cards. If you're going to upgrade your graphics card, you need to reduce the bottlenecks in the system.
why 3gb ram and not 4gb or 8gb++? at least have dual channel ram with 2 2gb sticks.
Because the defective Merom chipset in use in the Core2Duo systems did not support greater than a 4G memory mapping space, and 1G of that was taken up as I/O space, so it was unable to remap the extra 1G of physical RAM and.or move the I/O hoe, even though it had the physical address lines to do so.
The chipset was manufactured between Nov 2006 and Oct 2007, but was used far longer than that by many manufacturers, since Apple was soaking up almost the entire supply of the corrected chipset, which was manufactured between Nov 2007 and Oct 2009.
Intel screwed up, and then taped out anyway in order to meet market deadlines.
It typically wasn't a big deal for most people, since the 2G SIMMs were very unstable at that point, and even desktop systems rarely had more than 3 SIMM slots. This changed in 2009 when Hynix finally fixed their 2G SIMMs, but the company nearly bit the dust anyway, as by then it had defaulted on several loans and one debt-equity swap.
Most people only discovered the screwup in the Merom chipset that happened to be in their machine when they started trying to use 2 2G SIMMs in their Core2Duo machines with the old Merom, and were only seeing 3G of RAM show up to the OS.
It doesn't work like this though. Even if you take the 50% performance increase on face value (not taking into account higher AA/ASF/Shaders) that would mean a game running at 15fps would increase to 23fps. Not exactly much of an increase. Even if you were getting 30fps on the GTX 260, that's an increase of 15fps (which is what the tests essentially saw), hardly worth $300.
Meanwhile, if you spent the money on CPU/MBD/RAM & a mid range graphics card (say a GTX 480 at around $150), you'd see actual performance increases of around 3.5x that of sticking a GTX 660 on a crap motherboard with a crap processor.
Sure, if you had every intention of upgrading the rest of the components, the graphics card is going to be the easiest to swap out, but you're still going to need to upgrade the CPU/MBD/RAM.
The article hides the fact that the increase of a GTX 260 vs GTX 660 card in a modern system would be a ~400% increase in performance. Not sure what they're trying to prove, but to me it proves they know nothing about hardware, gaming or value for money.