NVIDIA recently released the somewhat entry level GTX 650, but it is performing at Radeon 7700 levels game performance wise. That means that there was a gap left open up-to the Radeon 7800 series. To fill that gap NVIDIA now is releasing one more product, it's the all new GeForce GTX 650 Ti (with that Tiindicating a distinct difference).
See the regular GTX 650 comes heavily castrated with just 384 shader cores whereas the Ti model is released with a much better 768 CUDA cores (shader processors) and a 925 MHz GPU clock (for the reference products). So that's nearly double the processing performance. The budget graphics card series for gamers therefore just got a little more interesting. Albeit we say that 150 USD/EUR should not exactly really carry the tag budget.
But if you do not want (or can not) to spend heaps of money on your HTPC card or simply game in resolutions lower then 1920x1080 then for the 150 bucks the GeForce GTX 650 Ti positions itself smack-down in-between AMD's Radeon HD 7770 and 7850, and that really is a comfortable spot to be in for many of you.
Being based on the Kepler GPU architecture yet obviously NVIDIA did not want to make yet another separate chip. So The GTX 650 Ti is the same one used on the regular GTX 660, yup the GTX 650 Ti is using the "GF106" silicon opposed to the GK107 being used on the regular 650 model (it's confusing as heck we know).
However that GK106 silicon is armed with 768 active shader processor cores divided over four processor clusters running at 930 MHz on the reference products. That means that the GeForce GTX 650 Ti packs decent enough punch whilst consuming a relatively small amount of power. Interesting to know is that the 650 series all have a fixed 3D clock, so there is no dynamic boost aka Turbo functionality or anything. Thanks to that it actually overclocks really well you will find out in our AIB/AIC reviews.
The board partners will released two models of the 650 Ti series, one with 1 GB and the other one 2GB of 5GHz GDDR5. The big bad limiter however will be the the memory bus width, it is 128-bit memory and that will have a seriously impact on the overall rendering and compute performance.
Overall the GeForce GTX 650 Ti, as we'll demonstrate, has enough horsepower to step into the DX11 games at up-to 1600x1200 resolutions, with the somewhat dated titles a resolution of 1920x1080/1200 will actually be playable as well. And if you can forfeit to medium quality settings in a game and don't do any crazy stuff Anti-aliasing wise, it's definitely plausible to play games really nicely at such resolutions versus acceptable framerates.
So in today's review we'll look at the 650 Ti from MSI, in specific the Power Edition. The SKU comes with a customized PCB, the Cyclone II cooler and in fact is factory overclocked towards 993 MHz on the core clock frequency. Have a peek at what is the MSI GeForce GTX 650 Ti power edition offers.
Reference technology and specs
So as always we'll first address the reference technology. A product like tested today is always based of the reference design and sometimes optimized a little in terms of VRM, PCB design, component usage, clock frequencies, cooling and display outputs. But the building template the manufacturers work from is in fact that reference design.
In this segment of the article we will look at the reference (original design) based specs and architecture. The GeForce GTX 650 Ti is based on the Kepler GPU architecture. The product however is using the GK106 which is based on a 28nm fabrication process. You guys all know the GK106 from the GeForce GTX 660 products. That chip has 2.64 Billion transistors.
The GeForce GTX 650 Ti whould be 40% faster than the GTX 650 on average.
When we peek at NVIDIA's slide decks we see them denoting that the GeForce GTX 650 Ti has been designed for 1080p gaming with medium settings and FXAA. That's true but please do understand this remains a budget level graphics card .
- The GeForce GTX 650 boasts 384 CUDA (shader) cores (GK107)
- The GeForce GTX 650 Ti boasts 768 CUDA (shader) cores (GK106)
- The GeForce GTX 660 boasts 960 CUDA (shader) cores (GK106)
- The GeForce GTX 660 Ti boasts 1344 CUDA (shader) cores (GK104)
- The GeForce GTX 670 boasts 1344 CUDA (shader) cores (GK104)
- The GeForce GTX 680 boasts 1536 CUDA (shader) cores (GK104)
The GeForce GTX 650 Ti ships with 4 activated SMX units containing 768 CUDA Cores and 64 texture units. The graphics core clock speed of the GeForce GTX 650 Ti is 925 Mhz (reference frequency - AIB/AIC partners will ship with higher clocks) . GPU Boost is not available. The GeForce GTX 650 Ti memory speed is loacked at an 5400MHz effective data rate based on a1350 MHz quad data rate for GDDR5.
You'll notice that the TDP of the card is set at maximum of 110 watts, however since a segment of the GPU is not used the realistic power consumption will be a little lower. We measured that to be roughly 90~100 Watts actually. So the GeForce GTX 650 Ti draws relatively little power. It still ships with an external power connector. This power connector provides additional headroom for overclocking. Many GeForce GTX 650 Ti boards are capable of hitting speeds in excess of 1100 MHz quite easily.
The idle power of GeForce GTX 650 Ti is 5W to 10W representing terrific in class idle power. In addition, HD video playable is ~13W, again representing best in class power consumption.
Display outputs include two dual-link DVIs and one mini-HDMI, however this could be different per board partner.
128-bit Memory Interface
The memory subsystem of the GeForce GTX 650 Ti consists of two 64-bit memory controllers in use (128-bit) with either 1GB or 2GB of GDDR5 memory.
The memory subsystem of the GeForce GTX 650 Ti consists of two 64-bit memory controllers in use (128-bit) with either 1GB or 2GB of GDDR5 memory.
We already mentioned it, another change is the ROP (raster operation) engine, cut down to 16 units opposed to 24 on it's bigger brother the GTX 660. With this release, NVIDIA now has the real mid-range products on its way. The new graphics adapters are of course DirectX 11.1 ready.
With Windows 8, 7 and Vista also being DX11 ready all we need are more new games to take advantage of DirectCompute, multi-threading, hardware tessellation and the latest shader 5.0 extensions. For your reference here's a quick overview of some of the more recent graphics cards.
| GeForce GTX 650 | GeForce GTX 650 Ti | GeForce GTX 660 | GeForce GTX 660 Ti | GeForce GTX 670 | GeForce GTX 680 | GeForce GTX 690 | |
| Stream (Shader) Processors | 384 | 768 | 960 | 1344 | 1344 | 1536 | 3072 |
| Core Clock (MHz) | 1058 | 925 | 980 | 915 | 915 | 1006 | 915 |
| Shader Clock (MHz) | 1058 | 925 | - | - | - | - | - |
| Boost clock (Mhz) | - | - | 1033 | 980 | 980 | 1058 | 1019 |
| Memory Clock (effective MHz) | 5000 | 5400 | 6008 | 6008 | 6008 | 6008 | 6008 |
| Memory amount | <2048 | <2048 | 2048 | 2048 | 2048 | 2048 | 4096 |
| Memory Interface | 128-bit | 128-bit | 192-bit | 192-bit | 256-bit | 256-bit | 256-bit |
| Memory Type | GDDR5 | GDDR5 | GDDR5 | GDDR5 | GDDR5 | GDDR5 | GDDR5 |
For Kepler, NVIDIA kept their memory controllers GDDR5 compatible. Memory wise NVIDIA has nice large memory volumes due to their architecture, we pass 2 GB as standard these days for most of NVIDIA's series 600 graphics cards in the high range spectrum. We expect most GTX 650 Ti cards to come with 1 GB of graphics memory in order to remain price competitive, the 2GB models will be roughlt 25 USD/EUR more expensive.