The GPU Flashback Archive series continues today with a recap of the NVIDIA GeForce 700 series, a series refresh which heralds part two of the Kepler family of GPUs. We can also remember it as a time when NVIDIA launched their first ever GTX Titan card and with it, a new pricing and retail strategy for truly high-end graphics card products. Let’s take a look at the new Kepler architecture GPUs, the cards that were popular with HWBOT members and some of the more memorable scores that have been posted since launch.
NVIDIA GeForce 700: Overview
The 2011-2013 period of history saw NVIDIA implement a more regular cadence to their high-end product launches and refreshes. One that saw the company launch a new GPU architecture every two years, with new product lines arriving each year. This means deriving two product lines per architecture with an improved version offered the second time out. This is what we saw with Fermi, an architecture whose potential was full realized at the second attempt. With the GeForce 700 series, which arrived proper in May 2013 with the arrival of both the GeForce GTX 780 and GTX 770, we have something different. The new cards arrived using a much bigger version of the Kepler architecture compared to what we saw on the NVIDIA 600 series.
With the launch of the GeForce 600 series we saw NVIDIA roll out a slimmer, leaner GPU designed more with PC gaming in mind than pure computation performance in GPGPU applications. The GTX 680 was based on the GK104, a Kepler architecture chip with 3.54 million transistors. The GeForce 700 series used a Kepler-based GK110, a much larger chip (561mm² vs 294mm²) with over 7 million transistors built with Compute performance as its primary function. The elephant in the room with this story however, is the arrival of a wholly new product type, the GeForce GTX Titan series. The first Titan card had arrived in February of 2013 boasting what many described simply as ‘Big Kepler’, the GK110 chip that eventually makes its way to the most popular 700 series card, the GeForce GTX 780.
The original GTX Titan card represented NVIDIA’s first step on creating a new income stream by selling a flagship card direct to consumers via its own, exclusive sales channel, just as it had previously done with QUADRO cards. The Titan series became, and continues to be one of the reasons that NVIDIA have been doing so well as a company, a cash cow that many enthusiasts today now resent. If you bought a GTX Titan in February of 2013 for $1,000 USD, you may been a bit aggrieved when the GTX 780 arrived just months later sporting the same GPU. Just this week we saw the Titan V rollout for $3,000 sporting ‘Big Volta’. It’s a business model that NVIDIA are still honing with each Titan launch, one that means that early adopters are paying more money for the ‘best-graphics-card-money-can-buy’ – and they’re paying it again and again.
The GeForce GTX Titan and the GTX 780 which arrived for $649 USD in May 2013 were in fact a little different from each other with more CUDA Cores and memory on the Titan card. Here’s a useful chart from Anandtech which summarizes the Flagship succession story of the era very nicely.
One thing that is apparent, the price of a flagship graphics cards has moved from being just under a $500 USD to $1,000. A progression that continues today with subsequent $2,000 USD and $3,000 USD price points. NVIDIA has clearly decided to follow Intel’s lead and grow its’ ‘prosumer’ catalogue offerings. The chart above reminds us that product segmentation remains a key strategy at every level. The GTX 780 most notably features 2,304 CUDA Cores, or Stream Processors compared to the 2,688 on the GTX Titan, which also has 32 more Texture Units. In reality NVIDIA shaved off 2 Stream Modules to keep GTX Titan owners feeling sufficiently superior.
Perhaps most notably in terms of Compute power, the Titan has double the amount of GDDR5 memory at 6GB. Both cards use a 384-bit memory bus with the GTX 780 tuned to have marginally higher frequencies on both the core and memory – a configuration best suited to gaming applications one would imagine. Both cards have a TDP of 250W, a good 55W higher than the previous Fermi-based- GTX 680. All GPUs are manufactured by TSMC with all 600 series and 700 series desktop chips using the 28nm process.
The GTX Titan used a shroud cooler similar to what NVIDIA developed for its GTX 690 card a year earlier. It’s trademark silver looks, grill and fan are in fact not so different from the Titan V today. The GTX Titan was pretty long at 10.5 inches (167mm), used a dual-slot design with 1x 6-pin + 1x 8-pin power connectors. In terms of outputs, we see some 2x DVI, 1x HDMI and 1x DisplayPort.
The Most Popular NVIDIA GeForce 700 Card: The GeForce GTX 780
It’s time to take a look at the most popular NVIDIA 700 series cards in terms of submission numbers to the HWBOT database. Although not technically a member of the GeForce 700 series family, we have included the GTX Titan in the data. As you can see, it was actually quite a rare beast in overclocking circles:
- -GeForce GTX 780 – 18.79%
- -GeForce GTX 770 – 17.43%
- -GeForce GTX 760 – 17.04%
- -GeForce GTX 780 Ti – 13.29%
- -GeForce GTX 750 Ti – 9.30%
- -GeForce GTX Titan – 4.07%
- -GeForce GTX 780M – 2.37%
- -GeForce GTX 750 – 1.97%
- -GeForce GT 740M (GK208) – 1.93%
- -GeForce GTX 765M – 1.93%
It’s perhaps interesting to note how the top three models are almost identically represented at the top of the table, just as we found with the previous generation. GTX 780 cards are most popular with 18.79% of all 700 series submissions on HWBOT historically speaking. The next more affordable rung on the NVIDIA GeForce ladder is the GTX 770 which finds itself used in 17.43% percent of all submissions, proving to a popular choice. As is the GTX 760 which has been used in 17.04% of submissions.
The NVIDIA reference design for the GTX 780 was in fact pretty much identical to that of the GTX Titan. Other vendors however were soon having fun creating their own interpretations. Here’s a three slot design from Inno3D for example.
Interestingly, 6GB versions of the GTX 780 did become available in September of 2013, but it is the GTX 780 Ti that arrives in November that we must crown as the fastest 700 series card. It was popular despite its $700 USD price tag and was used in 13.49% of submissions. This is followed by the more affordable Ti card, the GTX 750 Ti which garnered 9.30% of all submissions. All of these cards prove to be more popular than the GTX Titan which was used in just over 4% of all submissions.
Here’s an example of the GTX 760 card, in this case coming from MSI:
The GTX 780 Ti in fact had 2,880 CUDA Cores, 192 more than the GTX Titan card. It also had more Texture Units and higher memory and GPU clock frequencies. The only advantage a GTX Titan could offer is the larger 6GB memory buffer that makes it more attractive to those in need of a massive memory buffer..
Here’s an example of a GTX 780 Ti card from Zotac:
NVIDIA GeForce 700 Series: Record Scores
We can now take a look at some of the highest scores posted on HWBOT using an NVIDIA GeForce 780 Ti card, the fastest single GPU card in the 700 Series lineup.
Highest GPU Frequency
Although technically speaking, GPU frequency (as with CPU frequency) is not a true benchmark, it does remain an important metric for many overclockers. Looking through the database, we find that the submission with the highest GPU core frequency using a GeForce GTX 780 Ti card comes from the legendary k|ngp|n (US). He pushed the GPU of an EVGA GeForce GTX 780 Ti Classified card (which he probably helped design) to 2,051MHz, which is a staggering +127.38% beyond stock settings. The rig used also included an Intel Core i7 4960X ‘Ivy Bridge-E’ processor clocked at 6,170MHz (+71.39%).
You can find the submission from k|ngp|n here on HWBOT: http://hwbot.org/submission/2587809_kingpin_3dmark11___performance_geforce_gtx_780_ti_24714_marks
3DMark Vantage Performance
The highest 3DMark Vantage score submitted to HWBOT using a single NVIDIA GeForce 780 Ti card comes from Sofos1990 (Greece). He pushed a GTX 780 Ti to 1,960MHz (+117.29%) on the GPU and 2,100MHz (+39.81%) on the memory to hit a score of 84,470 marks. The rig also used a very nicely pushed Intel Core i7 5960X ‘Haswell-E’ processor clocked at 5,800MHz (+93.33%).
You can find the submission from Sofos1990 here on HWBOT: http://hwbot.org/submission/2722892_sofos1990_3dmark_vantage___performance_geforce_gtx_780_ti_84470_marks
The score was actually made during CES 2015 in Las Vegas.
In the classic Aquamark benchmark we find that Bullshooter (Germany) is the highest scorer with a single GeForce GTX 780 Ti card. He holds the Hardware First Place record with an impressive score of 575,741. The score was made with the help of an Intel Core i7 7700K ‘Kaby Lake’ chip clocked at 6,700MHz (+59.52%).
You can find the submission from Bullshooter here on HWBOT. From the wallpaper used it may well have been submitted as part of the GALAX GOC 2016 Qualifier contest: http://hwbot.org/submission/3439723_bullshooter_aquamark_geforce_gtx_780_ti_575741_marks
Thanks for joining us for this week’s episode of the GPU Flashback Archive series. Come back next week and check out the NVIDIA GeForce 900 series of graphics processors and cards.