The NVIDIA TITAN Z Performance Review
- Thread starter SKYMTL
- Start date
Conclusion
NVIDIA and I were at odds about this review from the very beginning. Frankly, their decision to avoid negative press by bypassing reviewers didn’t sit well at all. So I went at this review with a pitchfork, torch and angry mob mentality, figuring I would eviscerate the TITAN Z for being stupidly expensive, hot running, power hungry, loud and unable to compete on a level footing with AMD’s R9 295X2. While some of those points were proven true, for all my personal negative bias towards it, I have to admit the TITAN Z did turn out to be an extremely impressive graphics card.
Pushing aside that thousand pound gorilla called pricing for a moment, what NVIDIA has accomplished here needs to be commended. Without resorting to a complicated water cooling setup they have created a card that can outpace nearly everything else on the market without causing a racket or throttling when under long-term loading conditions. Miraculously, the TITAN Z runs quiet, is able to consistently reach higher-than-advertised Boost clocks and it still consumes significantly less power than AMD’s alternative. We couldn’t test overclocking headroom since our benefactor was afraid of damage to his $3000 card but supposedly there’s more headroom left untapped.
While sub-4K resolutions may not be the TITAN Z’s native territory, there’s no denying its ability to power through games like no-one’s business. There was a seesaw battle with the R9 295X2 for ultimate supremacy but NVIDIA’s flagship ended up eking out a lead of about 6% on average, which is no small thing considering the incredible frame processing horsepower of AMD’s card. The GTX 780 Ti SLI system reigns supreme at these resolutions due to its higher frequencies but bumping resolution to the next level results in a slightly different picture….
At 4K we see a narrowing of the field as the R9 295X2 is able to step up in a big way and leverage its large framebuffer to boost overall performance. The TITAN Z remains competitive and actually narrows the gap between itself and the GTX 780 Ti SLI setup by excelling in situations that leave two GTX 780 Ti’s grasping for memory capacity. Some excellent examples of this are in Far Cry 3, Assassin’s Creed Black Flag and Thief where the overall experience is noticeably smoother. Does this justify the $1000 premium over a pair of 6GB-totting TITAN Blacks? While I don’t have any of those on hand, let’s venture an “Absolutely Not” for now. And yet, these results will be reason enough to persuade some gamers to take the plunge.
Ultimately though the TITAN Z’s downfall may be the identity crisis it is currently going through. It has a product page dedicated to CUDA development and is priced accordingly, yet NVIDIA’s introductory video doesn’t mention professional uses even once (gaming is mentioned six times though). There’s a completely justified feeling in the enthusiast community that the TITAN brand’s slim feature set advantage is being used as an excuse justify charging outrageous prices for what is still a gaming-first card. Remember, this is a GeForce product so its roots are firmly planted in the gaming space rather than within any professional segment and putting on a dash of double precision lipstick doesn’t change that.
While I certainly can’t fault NVIDIA for charging with a capitalist market will bear; looking at the raw figures is sobering. The Z costs nearly as much as three TITAN Blacks but barely offers the performance of two GTX 780 Ti’s. AMD’s R9 295X2, a card that can compete with it on a level footing costs HALF as much. Some may argue that the TITAN Z’s value lies in an ability to fit into small form factor cases but, judging from the thermal buildup around the card itself, putting it into a confined chassis with limited airflow would be a recipe for disaster.
The term “value” can mean wildly different things to different people. However, the TITAN Z’s so-called value will be dubious for many in the double precision CUDA world as well. Most developers we’ve spoken to have the space, power and cooling capacity for a pair of TITAN Blacks, a combination which promises to one-up the Z’s performance in nearly every respect. To them, the only way NVIDIA’s $3000 monster makes sense is in a dual card system that needs four CUDA-powering cores but only has two or three compatible PCI-E slots.
With AMD launching the R9 295X2 at its current performance levels the TITAN Z feels like a tantalizing taste of what could have been an awesome competitor. Instead, NVIDIA had no way of convincingly beating AMD’s dual GPU card while remaining on air cooling so they likely threw in the towel and brandished the TITAN name once again. Whether or not we all agree with the turn-tail-and-overprice strategy doesn’t really matter; the TITAN Z will still likely sell out due to the raw power, ease of installation and brand recognition it brings to the table. That certainly doesn’t make it a “better” card than AMD’s R9 295X2 though, regardless of what a $3000 price tag would have people believe.
Conclusion
NVIDIA and I were at odds about this review from the very beginning. Frankly, their decision to avoid negative press by bypassing reviewers didn’t sit well at all. So I went at this review with a pitchfork, torch and angry mob mentality, figuring I would eviscerate the TITAN Z for being stupidly expensive, hot running, power hungry, loud and unable to compete on a level footing with AMD’s R9 295X2. While some of those points were proven true, for all my personal negative bias towards it, I have to admit the TITAN Z did turn out to be an extremely impressive graphics card.
Pushing aside that thousand pound gorilla called pricing for a moment, what NVIDIA has accomplished here needs to be commended. Without resorting to a complicated water cooling setup they have created a card that can outpace nearly everything else on the market without causing a racket or throttling when under long-term loading conditions. Miraculously, the TITAN Z runs quiet, is able to consistently reach higher-than-advertised Boost clocks and it still consumes significantly less power than AMD’s alternative. We couldn’t test overclocking headroom since our benefactor was afraid of damage to his $3000 card but supposedly there’s more headroom left untapped.
While sub-4K resolutions may not be the TITAN Z’s native territory, there’s no denying its ability to power through games like no-one’s business. There was a seesaw battle with the R9 295X2 for ultimate supremacy but NVIDIA’s flagship ended up eking out a lead of about 6% on average, which is no small thing considering the incredible frame processing horsepower of AMD’s card. The GTX 780 Ti SLI system reigns supreme at these resolutions due to its higher frequencies but bumping resolution to the next level results in a slightly different picture….
At 4K we see a narrowing of the field as the R9 295X2 is able to step up in a big way and leverage its large framebuffer to boost overall performance. The TITAN Z remains competitive and actually narrows the gap between itself and the GTX 780 Ti SLI setup by excelling in situations that leave two GTX 780 Ti’s grasping for memory capacity. Some excellent examples of this are in Far Cry 3, Assassin’s Creed Black Flag and Thief where the overall experience is noticeably smoother. Does this justify the $1000 premium over a pair of 6GB-totting TITAN Blacks? While I don’t have any of those on hand, let’s venture an “Absolutely Not” for now. And yet, these results will be reason enough to persuade some gamers to take the plunge.
Ultimately though the TITAN Z’s downfall may be the identity crisis it is currently going through. It has a product page dedicated to CUDA development and is priced accordingly, yet NVIDIA’s introductory video doesn’t mention professional uses even once (gaming is mentioned six times though). There’s a completely justified feeling in the enthusiast community that the TITAN brand’s slim feature set advantage is being used as an excuse justify charging outrageous prices for what is still a gaming-first card. Remember, this is a GeForce product so its roots are firmly planted in the gaming space rather than within any professional segment and putting on a dash of double precision lipstick doesn’t change that.
While I certainly can’t fault NVIDIA for charging with a capitalist market will bear; looking at the raw figures is sobering. The Z costs nearly as much as three TITAN Blacks but barely offers the performance of two GTX 780 Ti’s. AMD’s R9 295X2, a card that can compete with it on a level footing costs HALF as much. Some may argue that the TITAN Z’s value lies in an ability to fit into small form factor cases but, judging from the thermal buildup around the card itself, putting it into a confined chassis with limited airflow would be a recipe for disaster.
The term “value” can mean wildly different things to different people. However, the TITAN Z’s so-called value will be dubious for many in the double precision CUDA world as well. Most developers we’ve spoken to have the space, power and cooling capacity for a pair of TITAN Blacks, a combination which promises to one-up the Z’s performance in nearly every respect. To them, the only way NVIDIA’s $3000 monster makes sense is in a dual card system that needs four CUDA-powering cores but only has two or three compatible PCI-E slots.
With AMD launching the R9 295X2 at its current performance levels the TITAN Z feels like a tantalizing taste of what could have been an awesome competitor. Instead, NVIDIA had no way of convincingly beating AMD’s dual GPU card while remaining on air cooling so they likely threw in the towel and brandished the TITAN name once again. Whether or not we all agree with the turn-tail-and-overprice strategy doesn’t really matter; the TITAN Z will still likely sell out due to the raw power, ease of installation and brand recognition it brings to the table. That certainly doesn’t make it a “better” card than AMD’s R9 295X2 though, regardless of what a $3000 price tag would have people believe.
Last edited:
