GPU vs. CPU Upgrade

[JPB]

GPU vs. CPU Upgrade: Extensive Tests

CONCLUSION:

Changing the Generation of Graphics Card has More Benefits

Geforce 6 and 7 cards are hardly suitable for current games and modern LCD resolutions?the higher graphics settings in the test show that the older graphics chips really have reached their limits. This is obvious, at the very latest, when you try to combine a Geforce 6800 GT with a more powerful CPU, which does not achieve any palpable increase in 3D performance.

However, games are not 100% dependent on the graphics card; the Geforce 8 and 9 require a basic level of power, otherwise they are unable to exploit their 3D potential. The speed of the CPU should lie somewhere between 2600 and 3000 MHz; any lower, and the new graphics chips lose considerable performance.

There is no obvious advantage to quad cores over dual cores, at least according to the graphics-based benchmarks. In order for the Q6600 to compete with the dual core E6750, the same clocking rate is recommended. If you wish to combine an E2160 with a Geforce 8800 or Geforce 9, you will need to overclock. Without a clock rate of at least 2400 MHz, you will lose a considerable amount of graphics performance, because the card is not fully loaded.

The difference in performance among CPUs costing $77, $268 or even $1,237 (50, 170, 800 Euros) is actually relatively low. If you compare an E2160 at 1800 MHz to an E6750 or Q6600, you will find a 30% difference in the overall results. If the E2160 is overclocked to 2400 MHz, though, the difference in overall results is just 15%. The smaller cache of the E2160 budget CPU can be overcome by a higher clocking rate of up to 3 GHz.

The change to a new generation of graphics card achieves more, but the CPU should still have sufficient brawn to provide the basic level the card requires. Changing from a Geforce 6800 GT to a current Geforce 8800 or 9800 can quintuple the overall results for 3D games. Changing from a Geforce 7950 GT to one of the new G92 graphic chips will at least double the overall results.

If you convert the frame rate to percentages in order to filter out the weighting caused by high fps figures, it is possible to obtain an increase of over 1100% by changing from a Geforce 6800 GT to a current Geforce 8800 or 9800. If you change from a Geforce 7950 GT to one of the new G92 graphics chips, you can obtain a performance increase of up to 180%, with improved DirectX 10 effects. The maximum possible values will depend on the CPU?s performance level.

___________________


Since this mostly has to do with GPU I think. I posted it here in Graphics.

Really nice article.

 

[Foxery]

They did an awful lot of work to figure out that 6-12 month product cycles (GPU) produce greater advances than 2-4-year product cycles. (CPU) Snarkiness aside, the part about costs is key; spending twice as much on a video card gets you a lot more than spending twice as much on a CPU does. The difference between low and high end CPUs is silly when you consider the difference between budget and top-end GPUs.

 

[RussianSensation]

I thought we have reached these conclusions for the most part in early 2002-2003! In the last 5 years no less, the graphics card almost always provided a better performance improvement for gaming (under the assumption that it was paired with a modern cpu during its time). There are still games though that do benefit from a powerful CPU (microsoft flight sim X). Also, it'd be interesting to see a comparison of minimum framerates and not just average frames when testing for smooth playability.

 

[SolMiester]

Something wrong with their tests, in the Blacksite area tests, all of the Q6600 OC to 3.2 show a decrease in performance, yet every other chip shows improvement when clocked...

 

[JPB]

This article is from May 15th. So it is a little old. I did look around here however and didn't see it posted. So...here it is

 

[nRollo]

Originally posted by: JPB
This article is from May 15th. So it is a little old. I did look around here however and didn't see it posted. So...here it is

That's a good article, shows pretty clearly where to spend you money when buying a PC for gaming.

Personally I'm pretty happy you don't need to spend a ton of money on CPUs to get top gaming performance these days. I just switched form a QX6700@3GHz to a E8400@3GHz, don't even noitice the difference. (other than cost)

What would have been really nice in that article is if they would have included SLi results, or ATi cards as well with Crossfire results.

That would have given the whole picture for buying options and where money is best spent.
(e.g. E8400 & 2 x 9800GTX vs QX9650 & 1 X 9800GTX)

 

[chizow]

Guru3D's CPU Scaling Round-Up

Guru3D also did a similar comparison and came to similar conclusions. A few things of note from other reviews I've read, there's a pretty significant difference going from 1MB L2 to 2MB on the C2D family chips, but much less from 2MB to 4MB+. Also in a recent AT review of the 9800GTX or GX2 Derek found some serious CPU bottlenecking with Crysis and came to the conclusion that # of cores beyond 2 didn't matter, only core speed mattered in Crysis seeing gains up to 4GHz. It doesn't seem as if the other reviews show as much scaling.

The main benefits I've found in gaming with a Quad core is if you multi-task (games or apps) or if you record with FRAPs. Many modern games can utilize 75-80% of a fast dual core and with FRAPs recording you'll peg utilization and cause significant performance penalties. With a Quad its not an issue.

 

[biostud]

Repoxy