Dramatic Improvement Changing from Onboard to Dedicated Graphics Card

RE: A Demonstration of Dramatic Performance Improvement Changing from Onboard to Dedicated Graphics Card

Gentlemen, Question Mark!

Here's a convincing example of the performance enhancement of a budget system in which only change is using a dedicated graphics card in place of the integrated GPU.

After trying my mother’s next-door- neighbor- brewed computer > "Grandma's TurboKitten 3000" [Athlon II X3 450 @ 3.2GHz (3-core), 1X4GB DDR3 1600 RAM, Gigabyte M68MT-S2, GeForce 7025/ nForce 630a integrated graphics sharing 256MB system RAM, 420W PS, Win 7 Home], I was quickly convinced the video system was, shall we say as kindly as possible- absolute shite. Good ol’ mum doesn’t make any demands on this computer- I think she’s discovered 1D computing- that’s nothing but alternating thank you notes and angry letters to health care providers- but my Precision T5400 was inconveniently unplugged and I needed to look at a Sketchup model-an impossible task on the TK3000.

Running a free trial Passmark Performance Test 8.0 on TK3000, the benchmark rating was atrocious- an overall score of 262. In the baseline list of Athlon 450 computers, the top ranked has a score of 2159, [ASUS M5A 78L MB, and OCZ Vertex SSD, and significantly, a Radeon 6670] and TK3000 achieved the second lowest score for a an Athlon Ii 450 system.

While the CPU, memory, 2D, and disk scores were not miles worse than other systems with this CPU, the 3D score was the noticeably and dramatically awful. Where the top Athlon 450 system had a 2D / 3D of 542/1635 (with Radeon HD 6670), TK3000 scored 420 in 2D and an abysmal 24 in 3D ! The lowest rated system had a 3D score of 19! and sure enough, it was using the onboard GeForce 6159SE nforce 430, another, earlier incarnation of integrated graphics. The "n" in nForce must stand for "no" as the best performance in 3D for the 630a was 40. The 2D of 511 is not bad at all- I see megabuck Xeon zoomers with $700 Quadro 4000's that have 2D ratings in the high 400's- but they'll also have a 3D of 2200,.. Closer to home, my Dell Precision T5400 with a quad core Xeon 3.16GHz, 12GB RAM, and a Quadro FX 4800 (1.5GB) has a 2D score of 512 or =TurboKitten +1. However, the FX4800 3D score is 912 whereas the best 3D of the 7025 / 630a is 40- so take that noForce ! The worst 2D for the 7025 is 119 and the worst 3D is 4! Still, there are a few cases of 3D scores of 0,..

To make a long story short, I decided to find the best specification card I could for the lowest eBay Buy It Now price- so as not to spend forever on it- and quickly found a Newegg eBay offer of an EGVA GeForce GT 240,"manufacturer refurbished", for $30 and shipping included. The 128-bit GT 240 was never a brilliant device- about $100 in it's prime, but the later version includes attractive numbers > 1GB of GDDR5 memory (earlier were 512MB/GDDR3), 96 CUDA cores, a memory bandwidth of 54.4, and reasonable clock speeds, plus VGA, DVI, and HDMI ports. [FX 4800 1.5GB GDDR3, 192 cores and 78 mem b/w- and only $1,100 new! ] I liked the modesty of the 70W power rating of the GT 240 too. TheTK3000 has a 430W PS so it’s not at the power limit by any means and this card is probably not terribly hot-running. The need to watch graphic card power requirements- now I place it in the first tier of specifications to check- is relatively new to me. The previous card in the T5400 was a GeForce GTX 285 (1GB) and that took 205W and GTX 295's were 244. Lots of heat that needed consistent 1400rpm fans speeds. Fortunately, the T5400 has an 875W supply as Dell expected users to have two big Quadros, which can take a lot power, though Quadros are typically slightly detuned- clock speed and watts from the equivalent GeForce gaming version (FX4800 = GTX 260). It looks as though about all the high frame flier GTX cards grab a lot of electrons. A 680 takes about 200W and a 690 about 300W.

The GT 240 arrived from Newegg in only 4 days, and seemed to be a new card in an EGVA box. It’s a delicate little thing after handling the chunk-style GTX 285 and FX4800. I popped it in the sole PCI-x16 slot. As I’d never dealt with onboard graphics before, it occurred to me that I should check to see if it should be disabled in the way onboard audio is done, but I tried to get into the BIOS/ setup without success. I never saw a spin up config/setup prompt- what the hell key is it?. Plugged the 21" Acer into VGA as I didn’t have a DVI cable around, updated the driver through Control Panel > Device Manager and ran Performance Test 8.0 again.

Results were dramatically better. While the CPU. and disk numbers were improved slightly and the memory noticeably, I was stunned to see that the overall rating improved from 262 to 1394! - a factor of 5X. The significant numbers were the 2D/ 3D scores improving from 420 / 24 to 449 / 976. That is to say, the 3D score improved by a factor of 40X! In the Windows Experience Index, memory improved from 5.9 to 7.0. < Did the RAM-shared 256MB graphics memory cause a bottleneck?, Aero graphics went from 3.9 to 6.7, and the important 3D business/ gaming graphics went from 3.3 to 6.7. As the 3D score of 3.3 was the lowest parameter in the WEI, the overall score went from 3.3 to the new lowest- the 5.9 of the Disk, which everyone in the world must have who doesn't use an SSD, Amazingly, TK3000+GT240 became the fastest system using a GT240-bettering several Phenom II 955's on fancy smantzy ASUS X-MB’s! I’m not immune to wonderment / jealously since my super-dooper T5400 scored 1623 on Performance Test 8.0, not a million miles better than 1394 in %’s. Certainly, TK3000+GT240 gets the Most Improved Contestant Prize.

While those using their system for snerkling around the Web tubes, email, and letters may be comfortable with onboard graphics, I've seen some respectable rating results from the i7 integrated HD4000 ( better than some previous generation GeForce GTX, Radeon, and Quadros), and the onboard graphics of the upcoming Haswell CPU are supposed to be quite a bit better, I think this example, while using many words, tells a simple story. If you do anything with 3D, look at this example of the proportional improvement possible by changing from integrated to dedicated graphics after only 20 minutes' card hunting and $30!


4 answers Last reply
More about dramatic improvement changing onboard dedicated graphics card
  1. Thanks for sharing that information, bambiboom.
  2. Nice rundown. The results are exactly what they should have been.
  3. My brother uses his PC mostly to browse the web and watch movies, but I'm thinking about upgrading him to dedicated graphics anyways. If it improves other components like memory and CPU performance, it could be worth it. Looking for a very cheap card, I found a 4670 for 30$, someone selling it second hand nearby my location. If my brother is willing to pay for it I am curious to see if he'll notice general system performance improvement.
  4. MC_K7 said:
    My brother uses his PC mostly to browse the web and watch movies, but I'm thinking about upgrading him to dedicated graphics anyways. If it improves other components like memory and CPU performance, it could be worth it. Looking for a very cheap card, I found a 4670 for 30$, someone selling it second hand nearby my location. If my brother is willing to pay for it I am curious to see if he'll notice general system performance improvement.



    It's certainly worth looking into trying out a dedicated card in your brother's PC, and $30 to me is pretty close to free given the sophistication of the technology it's buying. Electronics is the only realm where it evolves to perform better for less and they sell really fairly recent and decent stuff for very little. There was a great cartoon- maybe in the New Yorker of a store table with some PC's piled around on it, labeled "Sale-Day Old". Certainly, the GT 240 was by far the most impressive $30 I'd ever spent on a computer.

    I'm still amazed at the differences the upgrade to dedicated graphics made, the most immediately noticeable was that the computer runs Sketchup really well instead of not at all. Another > part of the Passmark test a brief 3D animation scenes with fighter planes and animated water. With onboard graphics this test scene ran at 1-3 fps and after-up to 60fps! Beyond the 40X 3D change, I was also surprised at the collateral improvements, the CPU changed up a bit, but the the memory performance difference in WEI of 5.9 to 7.0 was notable. The shared RAM swapping to the graphics subsystem must make a bottleneck. Onboard graphics seems to represent the exact-opposite of the trend of freeing the CPU by sending more processing tasks to the GPU cores.

    Given the apparent level of use, my mum and your brother may not see anything improved after making the change, but then again, someday they'll never know they'd missed hitting that 3D brick wall!


Ask a new question

Read More

Graphics Cards Performance Graphics