What's new
  • Please do not post any links until you have 3 posts as they will automatically be rejected to prevent SPAM. Many words are also blocked due to being used in SPAM Messages. Thanks!



HardwareCanuck Review Editor
Staff member
Feb 26, 2007
We first talked about NVIDIA’s TITAN Z back when it was unveiled during the GPU Technology Conference in March. Since then, additional details have been non-existent but it goes without saying that during the last two months tweaking and no small amount of core binning was needed to bring the Z to market. The original launch date was actually April 29th so we’re looking at a month’s worth of delays for fine tuning. Now it is about to hit the channels in understandably limited quantities.

Now that launch day is finally here, we finally know a bit more about the TITAN Z and its potential capabilities. First and foremost the price has remained stable at a whopping $3000 which, for those keeping track at home, is enough coin to buy you two AMD R9 295X2 cards or three TITAN Blacks. However, this isn't a simple gaming card; it has much broader reaching implications.

Another mystery was clock speeds. With an air-based cooler and an insane amount of heat being generated by two fully unlocked GK110 cores and 12GB (yes you read that right) of GDDR5, some expected frequencies to suffer. Those predictions came true since things have been dialed down…way down.


In order to achieve their goals TDP and power goals NVIDIA had to lower core clocks to a point that’s under a GTX 780’s threshold. GPU Boost 2.0 brings clocks up to a reasonable level, matching a GTX TITAN’s output but actually achieving 876MHz on a continual basis may prove to be a challenge. Memory hits the 7Gbps mark so there’s no worry of any bandwidth bottleneck.

With all of these aspects taken into account, it only takes a bit of extrapolation to pinpoint where the TITAN Z sould end up in from competitive analysis standpoint. If we take the numbers achieved by AMD’s R9 295X2, GTX 780 Ti SLI and GTX 780 SLI in our recent comprehensive review, things don’t look all that great from a gaming perspective. With a best case scenario of about 15% lower core frequencies than the GTX 780 Ti, the $3000 Z would fall between the GTX 780 Ti SLI and GTX TITAN SLI setups at sub-4K resolutions. At 4K and frame buffer-limited scenarios there will likely be some scenarios where the 12GB of GDDR5 will win the day.

Those are some incredible performance numbers from a single PCB card but AMD has already thrown their hat into the ring with the R9 295X2. If we had to make a fearless prediction, the TITAN Z will (on average) lose to AMD’s monster at sub-4K resolutions. At 4K and above, the battle will likely end in a closer race since its insanely large memory size will only become telling factor in a few outlier, framebuffer-hogging games. This situation also points towards why NVIDIA’s board partners have been hesitant to launch premium GTX 780 Ti 6GB SKUs, despite the GTX 780 getting the 6GB treatment months ago.

Unfortunately we can’t test these theories since NVIDIA has strictly limited press coverage this time around, which is understandable considering the TITAN-Z isn't a predominantly gaming-focused card. The last thing they want is to have their $3000 darling losing to an alternate solution from the competition which costs half its price.


Let’s not allow potential gaming performance to misdirect us from the TITAN Z’s claimed target audience: CUDA developers. From their perspective, a triple slot solution that offers extreme amounts of double precision throughput will likely be a dream come true. To achieve TITAN-Z levels of performance they would have needed to look towards the Quadro lineup and spend upwards of $8000 to achieve a comparable workflow environment. So while many gamers may roll their eyes, GPGPU users will likely gobble these things up in no time.

This professional versus gamer-oriented situation presents us a bit of a quandary though. On one hand NVIDIA insists this is a card designed for the CUDA crowd, yet it carries the GeForce GTX brand, a name deeply-rooted in the gaming market. The TITAN Z really does swing both ways, but its driver development will continue down the GeForce “Game Ready” path and some enthusiasts will still pony up the three grand because they always do. If anything the TITAN series’ robust HPC backbone adds to the Z’s value in a big way, but gamers will be better served by less expensive alternatives despite the GeForce branding. CUDA developers on the other hand are busy screaming “take my money!”, and with good reason.

Before we cut this short, it should be mentioned that ASUS, Zotac and EVGA are NVIDIA’s launch partners for the TITAN-Z here in North America. For the time being ASUS and Zotac will stick to the reference design but EVGA has some pretty awesome looking custom versions coming out alongside their standard edition.


Foremost among them is the $3350 EVGA GTX TITAN Z Hydro Copper. With it, EVGA has teamed up with EK for the first time, creating an advanced slim-line full coverage water block which reduces the TITAN Z’s height from three slots to a more manageable two. This will come pre-overclocked at a base speed of 758MHz and can boost up to 941MHz. It may not come cheap but the HC will likely be the fastest kid on the block for 4K gaming.

In true EVGA fashion, there will also be an $3200 “SC” edition that uses a reference cooler but runs at 732MHz and 915MHz for the Base and Boost clocks respectively.

All in all this should be an interesting launch, especially when the reactions of gamers are taken into account. If you want to chime in about the TITAN Z, hit up the link to our forum comment thread above.
Last edited:

Latest posts