Nvidia Titan Xp

[HEXUS]

Nvidia's finest consumer GPU reviewed and rated.

Read more.

[LSG501]

And as usual Hexus completely ignores anything other than gaming for a gpu....

With the push towards gpu assisted processing and the fact that a large portion of people will likely be using this card (like it's predecessors) to do things like gpu rendering, encoding or other things where the gpu is faster, purely because it's a cheaper alternative to the tesla and quadro cards (which an 'equal' card is around 5k), it's about time Hexus actually started reviewing the cards in areas where they often get used such as the ones I mentioned above.

[Troopa]

junk,

[philehidiot]

And as usual Hexus completely ignores anything other than gaming for a gpu....

With the push towards gpu assisted processing and the fact that a large portion of people will likely be using this card (like it's predecessors) to do things like gpu rendering, encoding or other things where the gpu is faster, purely because it's a cheaper alternative to the tesla and quadro cards (which an 'equal' card is around 5k), it's about time Hexus actually started reviewing the cards in areas where they often get used such as the ones I mentioned above.

I think they're playing to their audience. They'll be working on their market research which will show gamers are their primary audience. There are different sites which cater for different markets which will review this card for their respective uses.

I think that they've investigated thermal throttling to the degree they have pretty much shows this thing is useless for most GPGPU, cryptocurrency mining, modelling, etc without significant modifications to the cooling system. This isn't the website for you if you're interested in non gaming applications. Perhaps Anandtech would be more useful for you?

[watercooled]

And as usual Hexus completely ignores anything other than gaming for a gpu....

With the push towards gpu assisted processing and the fact that a large portion of people will likely be using this card (like it's predecessors) to do things like gpu rendering, encoding or other things where the gpu is faster, purely because it's a cheaper alternative to the tesla and quadro cards (which an 'equal' card is around 5k), it's about time Hexus actually started reviewing the cards in areas where they often get used such as the ones I mentioned above.

I'm being pedantic, but GPUs aren't really well-suited to video encoding. Sure, they have hardware codecs on-die but their efficacy is inferior to software encoders like x264. The GPU itself i.e. the 'cores' aren't really well-suited to encoding either - it's one of those tasks that are just better-suited to big CPUs.

[LSG501]

I think they're playing to their audience. They'll be working on their market research which will show gamers are their primary audience. There are different sites which cater for different markets which will review this card for their respective uses.

I'm not asking for it to be just about how good it is with cuda but if you look at the market that this level of nvidia card is used in you'll see a lot of 'indie studios' or freelance professionals using it for things like CAD modelling (maya, 3ds max etc) and 3D rendering (vray-rt, octane etc). It is after all fundamentally a cheaper version of the quadro P6000 12GB (lacks ecc ram basically and about 1/5th of the price) and imo it should at least have a couple of tests to show how well it performs in the other areas it might be used in.

If this was the 1030 then obviously I wouldn't be expecting it to be tested in those apps but maybe testing it for things like media playback wouldn't go a miss as it might end up in a htpc for example.

Basically they need to expand the testing to include, at least at a basic level, some of the areas the card might be used in.

I'm being pedantic, but GPUs aren't really well-suited to video encoding. Sure, they have hardware codecs on-die but their efficacy is inferior to software encoders like x264. The GPU itself i.e. the 'cores' aren't really well-suited to encoding either - it's one of those tasks that are just better-suited to big CPUs.

Maybe encoding wasn't the right term in hindsight but adobe premiere and after effects can utilise the gpu (cuda in nvidia's case) quite heavily for processing the output if the card is supported (a titan will be)

[watercooled]

Yeah like I say I was being pedantic - they're hugely useful for various stages the video creation workflow, just the final encoding part isn't really one of them. It's like (well, almost identical to in practice) Intel's QuickSync - there are situations where it's an acceptable compromise e.g. if you need to encode real-time and your CPU cores are already busy, are power limited (hardware encoding should usually by more power efficient than software, depending on whether programmable hardware is also used and the extent of power gating on the processor of course) or... well that's all I can think of at the moment. No wait, another would be on the likes of the Atom line where the cores themselves aren't awfully fast for encoding. On a reasonably fast CPU, x264 can be simultaneously faster and achieve a better quality/ratio than QSV - you often see strange comparisons, comparing QSV speed to x264 on a medium preset which is comparing apples to oranges.

[DerbyDave]

and some people say Anandtech are slow with their reviews.

[Myss_tree]

And as usual Hexus completely ignores anything other than gaming for a gpu....

With the push towards gpu assisted processing and the fact that a large portion of people will likely be using this card (like it's predecessors) to do things like gpu rendering, encoding or other things where the gpu is faster, purely because it's a cheaper alternative to the tesla and quadro cards (which an 'equal' card is around 5k), it's about time Hexus actually started reviewing the cards in areas where they often get used such as the ones I mentioned above.

+1.
I skip the gaming pages and go from 3D mark to PTN page and then conclusion.
Of more interest to me would also be the inclusion of Bang for Buck and Bang for Watt.

[Tarinder]

Just to add some context:

Nvidia Titan Xp was not a mass-market card and was not seeded to press by Nvidia, unlike all of the other GeForce cards. This is why you see so comparatively few reviews for it - we had to procure it ourselves.

However, Titan Xp is still positioned as a gamer's card so we evaluated it that way, and a number of high-profile SIs sell it as such. We expect to see a few systems come in with watercooled Titan Xps, enabling the card to be pushed to the limit and comfortably outdistance GTX 1080 Ti OC.

As for the more 'relevant' apps, Nvidia has Quadro and Tesla for a reason, and I agree those should be tested in that manner. All we wanted you to see was whether it made sense for well-heeled gamers to look at the reference Titan Xp compared to GTX 1080 Ti OC and the inevitable upcoming Radeon RX Vega.

[watercooled]

It is essentially a gamer's card, I agree - there's really nothing outside of branding and price that separates it from the GeForce line like there once was with the original Kepler Titan and hence not really any more (or less) reason this should be evaluated as a 'professional' card than any of the rest of the GeForce line.

[kalniel]

I think they're playing to their audience. They'll be working on their market research which will show gamers are their primary audience. There are different sites which cater for different markets which will review this card for their respective uses.

I think that they've investigated thermal throttling to the degree they have pretty much shows this thing is useless for most GPGPU, cryptocurrency mining, modelling, etc without significant modifications to the cooling system. This isn't the website for you if you're interested in non gaming applications. Perhaps Anandtech would be more useful for you?

Very good point, though I think expressing comment about our interests is (one piece of) feedback that helps them judge their audience too

[LSG501]

As for the more 'relevant' apps, Nvidia has Quadro and Tesla for a reason, and I agree those should be tested in that manner. All we wanted you to see was whether it made sense for well-heeled gamers to look at the reference Titan Xp compared to GTX 1080 Ti OC and the inevitable upcoming Radeon RX Vega.

But as I said earlier, a lot of smaller companies, freelancers and even hobbyists will use a titan class gpu INSTEAD of the quadro/tesla because they are basically the same but cheaper.

While it may be marketed for gamers by Nvidia it's because they don't want to show that it's pretty much identical in performance to the M6000 quadro so they 'artificially segment' the quadro and geforce cards by price and marketing.

This card is being or will be used by a LOT of small companies and freelancers, just like the previous titans because a small company paying the £5000 for the quadro of 'equal performance' is just not viable but they can afford to buy the Titan Xp at roughly a 5th of the price. And then you also have the 'hobbyists' who buy this card for the best of both worlds so they can do their hobby of 3D modelling etc and still play games.

I still feel that on the 'top of the range' cards you should test ALL or at least a range of possible usage scenarios, not just focus on the gaming performance which to be blunt is pretty obvious what the outcome will be.... they're going to be 'high scores' when it's a top of the range card.

I'm pretty sure AMD will push the fact that Vega RX is really good at open-cl etc based on vega frontier reviews (which funnily enough hits the same price point as the titan Xp...) so you can't really just class a card 'marketed' for gaming as purely for gaming anymore because people use their pc's for more than games these days and as such reviews should consider that.

[Tunnah]

As for the more 'relevant' apps, Nvidia has Quadro and Tesla for a reason, and I agree those should be tested in that manner. All we wanted you to see was whether it made sense for well-heeled gamers to look at the reference Titan Xp compared to GTX 1080 Ti OC and the inevitable upcoming Radeon RX Vega.

But as I said earlier, a lot of smaller companies, freelancers and even hobbyists will use a titan class gpu INSTEAD of the quadro/tesla because they are basically the same but cheaper.

While it may be marketed for gamers by Nvidia it's because they don't want to show that it's pretty much identical in performance to the M6000 quadro so they 'artificially segment' the quadro and geforce cards by price and marketing.

This card is being or will be used by a LOT of small companies and freelancers, just like the previous titans because a small company paying the £5000 for the quadro of 'equal performance' is just not viable but they can afford to buy the Titan Xp at roughly a 5th of the price. And then you also have the 'hobbyists' who buy this card for the best of both worlds so they can do their hobby of 3D modelling etc and still play games.

I still feel that on the 'top of the range' cards you should test ALL or at least a range of possible usage scenarios, not just focus on the gaming performance which to be blunt is pretty obvious what the outcome will be.... they're going to be 'high scores' when it's a top of the range card.

I'm pretty sure AMD will push the fact that Vega RX is really good at open-cl etc based on vega frontier reviews (which funnily enough hits the same price point as the titan Xp...) so you can't really just class a card 'marketed' for gaming as purely for gaming anymore because people use their pc's for more than games these days and as such reviews should consider that.

Mate I understand what you're saying, but most the folks here at Hexus don't give a toss about the enterprise side of GPU usage. There's a ton of other websites that go into that. Sure Tarinder could do a review and compare it to the other compute cards, and show how, while this is an expensive card, it's a cheap alternative to the others.

But literally 6 people could give a toss

[philehidiot]

I think they're playing to their audience. They'll be working on their market research which will show gamers are their primary audience. There are different sites which cater for different markets which will review this card for their respective uses.

I'm not asking for it to be just about how good it is with cuda but if you look at the market that this level of nvidia card is used in you'll see a lot of 'indie studios' or freelance professionals using it for things like CAD modelling (maya, 3ds max etc) and 3D rendering (vray-rt, octane etc). It is after all fundamentally a cheaper version of the quadro P6000 12GB (lacks ecc ram basically and about 1/5th of the price) and imo it should at least have a couple of tests to show how well it performs in the other areas it might be used in.

If this was the 1030 then obviously I wouldn't be expecting it to be tested in those apps but maybe testing it for things like media playback wouldn't go a miss as it might end up in a htpc for example.

Basically they need to expand the testing to include, at least at a basic level, some of the areas the card might be used in.

Honestly I think it would be pointless as the testing in those areas would be so superficial as to be useless to anyone who is seriously considering this card for those kinds of workloads. If you're going to do it, it's better to do it properly. A couple of tests just to get a taster isn't going to influence anyone's buying decision. Plus it would mean developing testing protocols for a type of card which rarely makes it onto this site. It's surely better to devote the time and resources into assessing gaming prowess as best they can and leave other areas to other sites to do an in depth analysis on? Otherwise we're back to the bad old days of magazines where space was limited and so you'd end up with benchmarks from one game and a 300 word review which was neither use nor ornament.

I personally think this review is extremely useful as it shows there is just no point in a gamer spending the extra money on a Titan and that is only possible because of how thoroughly they've tested it in this use case. I'd much rather see that kind of in depth analysis rather than spreading themselves two thinly. Otherwise, if you try and cover all subjects and use cases in detail you end up going down the route of Anandtech where people who can do reviews to the level you need are few and far between so reviews take ages and are often too late to be useful. Then you become and interesting site to read, but not one which people can rely on to make their next purchase.

The other thing is that there will always be individuals who want more from a review site. If they listened to everyone who said "I want to see this" then they'd spend all their time meddling with previous reviews to add things in. They do have to stand relatively firmly on their little patch of ground and make sure they aren't encouraged to creep in every direction which is suggested as that's just a road to disaster.

[Lepnik]

Hmmmm.... And the winner is 1080 Ti. Okay.