I thought I would try and make a discussion on a subject which bothers me...
Please tell me if I'm doing something wrong, or if someone has already posted something on this..
Okay, so I have an interest in GPUs of course, and I decided to buy a few low-end cards and test them against each other with a price budget of £50 maximum for each card. There is a lot of confusion I think in the low end, especially with GeForce cards, where there exist many variants of the same card powered by different GPUs.
And example of this is the recently released GT 730. There exist three variants of this card according to the Nvidia website known simply as:GT 730 DDR3 128-bitGT 730 DDR3 64-bitGT 730 GDDR5
As far as I can tell from listed specs, the 64-bit DDR3 and GDDR5 variants use the GK208 core, whilst the DDR3 128-bit uses the older GF108 core. I went and purchased a GT 730 DDR3 128-bit to see for myself. The price range here for this card is £35-50.
As you see there is no clear indication aside from technical specs on the side of the box (and the use of DDR3 which narrows it to one of two) as to which variant this is.
I took the cooler off to confirm the GPU:
GF108-300-A1, the exact same core used in the GT 430, is used in this card. Even the clocks are the same as the GT 430 so it is a rebrand. What I find interesting is the date code says that the chip was made on the 11th week of this year, so Nvidia are clearly still making this chip after 4 years since its first release. (you may also note the "NV430" marking on the PCB)
With the GK208 being smaller 86mm^2 vs 116mm^2, and more power efficient as well as quicker (which I test later), why is this necessary?
To underscore my point, I purchased a GT 630 (which I note is the Rev. 2.0 variant using the GK208 core and DDR3 memory). The card was only £35.
I apologise for the poor image quality, this was taken from my tablet.
As you can see there is no clear indication of which variant this card is.
I then tested both cards with three benchmarks to determine their performance. Product numbering alone would make one think that the GT 730 was faster than the GT 630 but that is not what my testing revealed.
The settings for each benchmark are:GT 730 DDR3 128-bit
Core clock: 700 MHz (1400 MHz shaders)
Memory Clock: 700 MHz (1400 MHz effective) DDR3
Memory Bandwidth: 22.4 GB/s
OC Core clock: 850 MHz ( 1700 MHz shaders)
OC Memory Clock: 900 MHz (1800 MHz effective)
OC Memory Bandwidth: 28.8 GB/sUnigine Sanctuary 2.3
Score: 1837Unigine Valley 1.0
Score: 1295Unigine Heaven 4.0
Score: 620GT 630 Rev. 2.0
GPU: GK208-???-?? (I believe it is GK208-301-A1 but I haven't taken the cooler off yet)
Core clock: 902MHz
Memory Clock: 900 MHz (1800 MHz effective) DDR3
Memory Bandwidth: 14.4 GB/s
OC Core clock: 1050 MHz
OC Memory Clock: 1100 MHz (2200 MHz effective)
OC Memory Bandwidth: 17.6 GB/sUnigine Sanctuary 2.3
Score: 2218Unigine Valley 1.0
Score: 1620Unigine Heaven 4.0
It is apparent, that even with almost half the memory bandwidth, the GT 630 is faster than the GT 730. I also note that the GT 730 overclocked cannot even match the GT 630's stock score. I know that my GT 730 is not the "highest end" variant, but I wanted to make a point. To the average consumer who knows little about the internal workings or specs (with all of these variants being available for sale) they would assume the GT 730 to be the faster card. It is not. In addition, the GT 630 was nearly £15 cheaper than my GT 730, so consumers buying by "higher price means more performance" will also lose out here.
So this is my first attempt to be constructive on The Tech Report forums, I hope you like it
What do you think of this matter?