Modern computing’s direction is definitely heading for consolidation and as for the consumer desktop market, a lot of the core functions that used to be performed by multiple controllers are now located on-die, in the CPU itself.
Still, it remains to be seen when the CPU and GPU will ever be married into a single package. Intel has been hard at work trying to improve their integrated graphics solutions but even with these significant improvements, they are still struggling to deliver a solid visual experience. The same effort goes out to their marketing in making sure they let people know they have a good enough IGP. But let’s admit it, people that worry about graphics are mostly the type that actually care what they use in their systems so much more so for people who actually play modern 3D games. It’s still the same old story.
Refer to the chart below for comparative of Intel HD4600 against NVIDIA GeForce GT 620 and the new GT 630:
The chart above clearly shows the performance gap that IGP and a discrete GPU, particularly the new NVIDIA GeForce GT 630 peaking at 360% and averages 157% across the charts.
Visually, the integrated graphics on the Intel processor also do not have enough details for a good gaming experience. Notice the screenshot above and a closer look will show that a lot of details are missing from the IGP but the game played on a NVIDIA GeForce GT 630 has much more detailed in-game textures, shadows and lighting.
Games aren’t the only applications to benefit from a dedicated GPU. Many modern applications we use on a day-to-day task leverage GPU acceleration, utilizing the increased NVIDIA CUDA core count to perform simple tasks.
Refer to the chart below for a rundown of such applications.
Again we see how dedicated GPUs from NVIDIA show much improved performance than the competing IGP from Intel. You may argue though that even with all these presented, choosing to go with an IGP will save you money. Well not exactly. Refer to the table below:
| Dedicated Memory (MB) | Memory Bandwidth (Mhz) | Platform Cost (USD) | |
|---|---|---|---|
| Haswell i5/i7 iGPU | 128 | 800 | $293 |
| GeForce GT 620 | 1024 | 1800 | $257 |
| GeForce GT 630 | 2048 | 1800 | $272 |
The table illustrates the amount of dedicated memory we get from all solutions. NVIDIA graphics card come with dedicated memory 2GB for the GT 630 and 1GB for the GT 620, both clocked at 1800Mhz . Intel’s latest IGP only has 128MB dedicated and borrows memory from the system which could be anywhere between 512MB to 1GB which is a lot already and can really hurt the performance of the system causing unresponsiveness and slowdowns. NVIDIA‘s discrete GPU’s also have faster memory and offers more than twice the bandwidth of the IGP solution.
In the table earlier we also show the platform cost of both solutions, breaking down the price at which consumers can acquire the systems tested:
Even with less parts, the Haswell system is easily the most expensive proposition. We’ve selected the lowest priced options we can find online, in the case of the processor, the Haswell i5-4430 is the lowest model to have the HD4600 IGP. On the other hand, the 3rd-generation Core i3-3320 paired with the NVIDIA discrete GPUs have a lower acquisition costs but as our charts indicate: NVIDIA Discrete GPUs are the smart choice. The GeForce GT 630 GPU beats the Haswell i5/i7 IGP, in most instances even the GeForce GT 620 also beats it in gaming applications. For multimedia and daily applications, both the GT 630 and GT 620 beat the Haswell chip showcasing 7x faster performance. All of these at a fraction of the price of a Haswell system.
Right now is not the best time to get a new CPU. A better return of investment can be gotten from upgrading your GPU.
We’ve seen in the test about what the new NVIDIA GeForce GT 630 can do and why it is the best choice, let’s take a closer look at the GeForce GT 630 64-bit.
New Kepler Architecture & New Process (28nm)
- More CUDA Cores (96 vs 384 Cores)
- More Performance (+15% Perf)
- Energy Saving (65W->25W, Save 40%)
- Support Multi-Display & 4K Display
A quick comparison with the older Fermi model, will show us the actual improvement the card has:
From the specs sheet above, the new GeForce GT 630 is a lot better than the older generation model. The NVIDIA® CUDA® core count has increased from 96 to four times more at 384 for the new Kepler model. The most notable and definitely interesting improvement is the power draw which at only 25w peak, is just incredible. To convince us further, the chart that follows shows us the performance improvement of the new GeForce GT 630 based on the Kepler architechture:
The newer GeForce GT 630 shows an average improvement of around 14% with some games showing nearly a 40% improvement over the previous generation GT 630.
As mentioned earlier, upgrading the CPU isn’t the best course of action if you want a boost in gaming performance in your system. If you have a system that’s about 5-years old, a GPU upgrade could be the better option for you. If you’re building a low-power day-to-day PC, the system we used for our discrete GPU testing is good starting point. With a dedicated GeForce GT 630 based on the Kepler architecture, you’ll get a 57% performance increase over integrated solutions while keeping the system responsive with 1GB~2GB dedicated memory which offers 2x memory bandwidth.
Play better games, have faster applications, more stability in your system at a more affordable range with NVIDIA GeForce GT.