NVIDIA GeForce GTX 780 Ti 3 GB

[haswrong]

I honestly don't know what kind of use 6GB would have on a card like this..

depends on the renderer. try thinking outside of the box a bit, and assume somebodys going to introduce a realtime raytracer.. then you shall utilize it and maybe need some more. i understand that vram has grown in size quite a lot for past few years, i started with 2 megs of ram, now have 512 of it. but that doesnt mean it is now balanced for next redering solution.

 

[Nihilus]

0

 

[Nihilus]

Sorry bro .. AMD is not nVidia ...
you get that price down only bcuz AMD are good guys !
without AMD your fan boys still bought titan for 1k$ and maybe ever higher.
1,000 USD Demo version ... joke of the year edition.

So true, this is the HD 4870/4850 all over again. Look how much the overcharged for the gtx 260 until AMD came in. As for the gtx 780ti, just another overpriced e-peen card. It does nothing useful in single configuration over the 290x in resolutions below 4k rez except 5 frames here and there, and then it nose dives in 4k when paired in SLI (see bf4).
Yeah, yeah, the 290x has a crap cooler, but that is a much easier fix than a $700 price tag and 3gb memory.

 

[Xzibit]

Are there any other reviews that detail throttling ?

Tom's Hardware noted you had to have the reference cooler running at 80% to stop thermal throttling in Metro Last Light to avoid it for a 30min gaming session.
Nvidia GeForce GTX 780 Ti Review: GK110, Fully Unlocked - Fan and Video Comparison

PcPer did similar thing with the 290X and found that it needed 50%+ fan speed to avoid throttling.

AMD & Nvidia are benchmark-marketing these cards even more now.

I don't play benchmarks but I do like to play the occasional game here and there and right now its BF4. Gaming sessions don't last 3mins either so are any site doing these test yet ?

 

[HumanSmoke]

Are there any other reviews that detail throttling ?

It would depend on what environment the card was tested in - and possibly ambient temp, although most review sites are European and North American so the latter shouldn't be an issue.
Hardware France - whose reviews I trust more than TH, didn't note any significant throttling

The GeForce GTX 780 operates between 863 and 915 MHz, with an average of 876 MHz, against a maximum frequency of 1006 MHz. As for the GTX 780 Ti, its frequency varies between 876 and 967 MHz for an average of 915 MHz, whereas our sample can rise up to 1019 MHz.

915MHz average sits fairly close to the minimum guaranteed boost of 928 taking into account a fluctuating workload. That may change with the non-default +106% setting, but then, the R9-290X's "Uber" setting isn't the default either...which just goes to show how quickly conventional benchmarking comparisons are becoming irrelevant.

 

[Xzibit]

It would depend on what environment the card was tested in - and possibly ambient temp, although most review sites are European and North American so the latter shouldn't be an issue.
Hardware France - whose reviews I trust more than TH, didn't note any significant throttling

915MHz average sits fairly close to the minimum guaranteed boost of 928 taking into account a fluctuating workload. That may change with the non-default +106% setting, but then, the R9-290X's "Uber" setting isn't the default either...which just goes to show how quickly conventional benchmarking comparisons are becoming irrelevant.

I don't trust one source either.

PcPer didn't mention which Nvidia card they used in the 290X test but in the long time comparison they showed the Nvidia boost clock dropped and never went back up again.

What i'm currently understanding is you need to run 290X @ 60% fan speed & 780 Ti @ 80% fan speed to keep throttling in check. From these limited examples and increasing fan speed will increase power consumption and noise.

With the intro of PowerTune and GPU Boost we might just be looking at top clocks being reached only at game intros and would have to manually adjust settings to reach results.

I'm more looking for long term time throttling/clock test per game. At the minimal like what Toms Hardware did. A benchmark loop for a longer duration to see if performance is decreased or sustained.

 

[Bjorn_Of_Iceland]

I can't think of any console port that needs this card. Not for that price anyways.

CoD Ghosts needs all that power for its dog A.I.

 

[W1zzard]

It would depend on what environment the card was tested in - and possibly ambient temp, although most review sites are European and North American so the latter shouldn't be an issue.
Hardware France - whose reviews I trust more than TH, didn't note any significant throttling


915MHz average sits fairly close to the minimum guaranteed boost of 928 taking into account a fluctuating workload. That may change with the non-default +106% setting, but then, the R9-290X's "Uber" setting isn't the default either...which just goes to show how quickly conventional benchmarking comparisons are becoming irrelevant.

I did a statistical clock speed analysis on page 31:

as you can see it does go down, but average and median are very high

 

[the54thvoid]

Are there any other reviews that detail throttling ?

Tom's Hardware noted you had to have the reference cooler running at 80% to stop thermal throttling in Metro Last Light to avoid it for a 30min gaming session.
Nvidia GeForce GTX 780 Ti Review: GK110, Fully Unlocked - Fan and Video Comparison

PcPer did similar thing with the 290X and found that it needed 50%+ fan speed to avoid throttling.

AMD & Nvidia are benchmark-marketing these cards even more now.

I don't play benchmarks but I do like to play the occasional game here and there and right now its BF4. Gaming sessions don't last 3mins either so are any site doing these test yet ?

I'm pretty sure it will throttle (all boost 2.0 cards do). My Titan could hit 993 at stock (no OC) but after a few mins it would start to downclock to keep the temp at 79 degrees.

However, the difference is that the Kepler cards advertise a base clock and base boost whereas the actual boost is far higher. In advertising terms, the Kepler card will generally remain above the marketed 'base boost' value.
Hawaii on the other hand is being 'touted' as a 1000Mhz card so when it down throttles it seems to be under performing. I know AMD have not stated anywhere it is guaranteed to run at that speed but i think it's just a miscommunication on their (PR) part that the cards are throttling badly. It's more a case they are working as designed but the thermals wont allow them to hit the 'advertised' boost. What does the actual retail packaging say for a 290X?

 

[alwayssts]

Very well done review, as always.

I think I take away from this what most people do:

The AMD cards will be very similar when you clock them to a similar level with a decent cooler that won't throttle (as badly). Sure, you might need ~1050mhz and slightly more than 5000mhz ram on 290x to compete with this at stock, and ~1200mhz/>6000 or so to compete with it overclocked, but who says that isn't possible, even if it does use more power? The ram is rated to 6ghz, which should allow the clock to scale to that level in a non-gimped (by bios, voltage, or whatever other nonsense comes about) configuration, and the voltage scaling in the 290x article clearly shows this as a possibility...perhaps hitting 1200mhz right around that 1.263 level ATi always seems to unofficially target.

Someone said 'bring on 290x Ghz ed'. In one capacity or another (officially or through partner cards), I don't think that is far from the truth...What AMD needs I've pretty much outlined above, not to say that others haven't probably reached the same conclusion.

I agree with Xzibit whom said they are both evermore making 'benchmark review' cards, and often it appears obvious they are targeting certain limitations of the competition's configuration (like AMD's 5ghz ram clock in this case, which would bottleneck it under the 780ti median boost clock). It would appear that TPU is a valuable source to their analysis teams, as the ~1% average wins (across this suite) and average/median clocks vs the main competitors specs perpetually show this to be the case.

I think you should feel loved, W1z.

 

[HumanSmoke]

What does the actual retail packaging say for a 290X?

Well, truth is obviously the first casualty in XFX and MSI advertising...although there's probably a disclaimer somewhere on the site that notes that the fan must be run at 100% in a walk-in freezer.

You can add Asus and Gigabyte to that list. Sapphire and HIS ( "up to 1000MHz") and Club3D ("~1000MHz") seem a little more circumspect.

 

[alwayssts]

However, the difference is that the Kepler cards advertise a base clock and base boost whereas the actual boost is far higher. In advertising terms, the Kepler card will generally remain above the marketed 'base boost' value.

Correct. It's a marketing fallacy.

Look back at W1z' 600/700 series reviews. While cards are marketed as x base clock and x+1 boost clock (median), typical speeds usually vary from x+2 to x+3 (consistantly across most boards from varying aibs for the same product).

I don't know if it is them being conservative so they don't look like amd does now (and arguably they did with the 'up to 1112mhz' 680 that ran around 1085mhz), to throw off the competition from the real clock (which is usually a bin or two higher than the difference listed), or simply to appear to have more performance at a lower clock than they actually do.

For instance, the only somewhat-honest comparison you can make here is that a (993-)1020mhz 780ti is ~15% faster than a 'usually' 947mhz 290, as those are consistent values for the most part.

Whatever way you look at it, it's frustrating. I personally find it deceiving, but understand how others may see it differently.

 

[Xzibit]

I'm pretty sure it will throttle (all boost 2.0 cards do). My Titan could hit 993 at stock (no OC) but after a few mins it would start to downclock to keep the temp at 79 degrees.

However, the difference is that the Kepler cards advertise a base clock and base boost whereas the actual boost is far higher. In advertising terms, the Kepler card will generally remain above the marketed 'base boost' value.
Hawaii on the other hand is being 'touted' as a 1000Mhz card so when it down throttles it seems to be under performing. I know AMD have not stated anywhere it is guaranteed to run at that speed but i think it's just a miscommunication on their (PR) part that the cards are throttling badly. It's more a case they are working as designed but the thermals wont allow them to hit the 'advertised' boost. What does the actual retail packaging say for a 290X?

What I would like to know is if these cards are even capable of delivering there clocks in a gaming run.

Benchmarking well we have the answers. What we don't know is duration and settings needed to maintain the performance.

It would be time consuming. Same benchmarks and loop it for 30mins to see if the clocks can be maintained. In this case it will be thermal limits to maintain it from down clocking so it will vary by card.

I cant use any other examples since no one has bothered yet.

COLD RUN

WARM RUN

The Run 2 graph shows the same 40% fan speed we results for the 290X on the previous pages but it also shows how the GeForce GTX GPU reacts. The NVIDIA card also drops in clock speed (from about 1006 MHz to 966 MHz, 4%) but it maintains that level throughout. That frequency is also above the advertised base clock.

Doesn't mention maintaining boost clocks
Its hard to tell since he didn't disclose what card he was using.

COLD RUN

WARM RUN

^This is only Crysis 3 60sec run @ 2560x1440

Metro Last Light Max Settings 30minutes Gaming Loop

^Now if this can be done per game/resolution/settings to find out throttling thresholds per cards to add to the benchmarks results. It would serve a site well. Then you would know at the very least in a control environment such as a open air test bench an idea of what you can expect to do once you purchase one of these to maintain the performance in gaming. Not just a quick bench run less then 3mins but a 30min gaming experience.

That's why I asked If anyone knew any sites doing such test.

 

[Deleted member 6693]

Does the 780 TI support DirectX 11.2?

Nope - just found out:
GTX 780 Ti GPU Engine Specs:
2880CUDA Cores
875Base Clock (MHz)
928Boost Clock (MHz)
210Texture Fill Rate (GigaTexels/sec)
GTX 780 Ti Memory Specs:
7.0 GbpsMemory Clock
3072 MBStandard Memory Config
GDDR5Memory Interface
384-bitMemory Interface Width
336Memory Bandwidth (GB/sec)
GTX 780 Ti Support:
4.3OpenGL
PCI Express 3.0 Bus Support
YesCertified for Windows 7, Windows 8, Windows Vista or Windows XP
GPU Boost 2.0, 3D Vision, CUDA, DirectX 11, PhysX, TXAA, Adaptive VSync, FXAA, NVIDIA Surround, SLI-readySupported Technologies
Yes3D Vision Ready
Microsoft DirectX 11.1 API

 

[sweet]

Well, truth is obviously the first casualty in XFX and MSI advertising...although there's probably a disclaimer somewhere on the site that notes that the fan must be run at 100% in a walk-in freezer.
http://img.techpowerup.org/131108/2...aphics_Cards/R9290X4GD5/#specifications"]Asus and Gigabyte to that list. Sapphire and HIS ( "up to 1000MHz") and Club3D ("~1000MHz") seem a little more circumspect.

LOL, choose higher fan target and your card will stable at 1000 MHz. Powercolor did the right thing when removing the quiet bios
http://www.techpowerup.com/reviews/Powercolor/R9_290X_OC/31.html

Just choose uber mode and you will be fine. If you can't live with that noise, 40 bucks for an after market cooler with help you, like in my case.

 

[qubit]

Does the 780 TI support DirectX 11.2?

Nope - just found out:
GTX 780 Ti GPU Engine Specs:
2880CUDA Cores
875Base Clock (MHz)
928Boost Clock (MHz)
210Texture Fill Rate (GigaTexels/sec)
GTX 780 Ti Memory Specs:
7.0 GbpsMemory Clock
3072 MBStandard Memory Config
GDDR5Memory Interface
384-bitMemory Interface Width
336Memory Bandwidth (GB/sec)
GTX 780 Ti Support:
4.3OpenGL
PCI Express 3.0 Bus Support
YesCertified for Windows 7, Windows 8, Windows Vista or Windows XP
GPU Boost 2.0, 3D Vision, CUDA, DirectX 11, PhysX, TXAA, Adaptive VSync, FXAA, NVIDIA Surround, SLI-readySupported Technologies
Yes3D Vision Ready
Microsoft DirectX 11.1 API

Oh well, that scuppers any speculation that the B revision GK110 is for DX11.2 support.

 

[EarthDog]

Who was speculating that???!

 

[The Von Matrices]

Oh well, that scuppers any speculation that the B revision GK110 is for DX11.2 support.

Do you realize how confusing that would be? GTX 780 Ti would be a guaranteed B revision but GTX Titan and GTX 780 would be a mix of A and B. Then you'd have some retailers sorting A and B cards and marking up the B cards for a premium price. I know this is done to an extent with revisions of CPUs for overclocking ability, but very few retailers ever do this, and as far as I know there have never been major feature changes on different silicon revisions, only minor errata fixes.

LOL, choose higher fan target and your card will stable at 1000 MHz. Powercolor did the right thing when removing the quiet bios
http://www.techpowerup.com/reviews/Powercolor/R9_290X_OC/31.html

Just choose uber mode and you will be fine. If you can't live with that noise, 40 bucks for an after market cooler with help you, like in my case.

Completely out of curiosity, what heatsink did you use that only cost $40? The Accelero Xtreme III that everyone is recommending starts at $75.

 

[qubit]

Who was speculating that???!

Do you realize how confusing that would be? GTX 780 Ti would be a guaranteed B revision but GTX Titan and GTX 780 would be a mix of A and B. Then you'd have some retailers sorting A and B cards and marking up the B cards for a premium price. I know this is done to an extent with revisions of CPUs for overclocking ability, but very few retailers ever do this, and as far as I know there have never been major feature changes on different silicon revisions, only minor errata fixes.

lol, no one in particular. I was simply saying just in case anyone was speculating about this, it scuppers it, that's all.

 

[EarthDog]

Only muppets speculate that.

 

[BiggieShady]

Only people who don't know that revision is called revision because it does not introduce new functionality, speculate that.

ftfy

 

[the54thvoid]

^Now if this can be done per game/resolution/settings to find out throttling thresholds per cards to add to the benchmarks results. It would serve a site well. Then you would know at the very least in a control environment such as a open air test bench an idea of what you can expect to do once you purchase one of these to maintain the performance in gaming. Not just a quick bench run less then 3mins but a 30min gaming experience.

That's why I asked If anyone knew any sites doing such test.

I know where you're coming from. From learning all about throttling from the Titan stock BIOS, lots of guys simply post up Afterburner logs. For a review site to do this would not be that intensive, AB logs aren't that large. Even screen shots post test run would bring up an indication of throttling.

My card throttled (on all stock settings) from heat. I only needed a small bump in fan parameters to negate it to be honest. But, I live in Scotland where my room temp is never above 20-23 degrees (even in our "Summer" ).

The other throttling from power is also present and that requires BIOS flashing - bad Nvidia

Do you realize how confusing that would be? GTX 780 Ti would be a guaranteed B revision but GTX Titan and GTX 780 would be a mix of A and B. Then you'd have some retailers sorting A and B cards and marking up the B cards for a premium price. I know this is done to an extent with revisions of CPUs for overclocking ability, but very few retailers ever do this, and as far as I know there have never been major feature changes on different silicon revisions, only minor errata fixes.

FWIW, this is from the Tech Report review.

When I asked Nvidia where it found the dark magic to achieve this feat, the answer was more complex than expected. For one thing, this card is based on a new revision of the GK110, the GK110B (or it is GK110b? GK110-B?). The primary benefit of the GK110B is higher yields, or more good chips per wafer. Nvidia quietly rolled out the GK110B back in August aboard GTX 780 and Titan cards, so it's not unique to the 780 Ti

ftfy

looooooooooool. I had to google "ftfy". I thought it meant..

Spoiler

fuck that fuck you

What can I say? I'm Scottish, swearing is our native tongue.

 

[radrok]

What can I say? I'm Scottish, swearing is our native tongue.

http://wiki.teamfortress.com/wiki/Demoman

depends on the renderer. try thinking outside of the box a bit, and assume somebodys going to introduce a realtime raytracer.. then you shall utilize it and maybe need some more. i understand that vram has grown in size quite a lot for past few years, i started with 2 megs of ram, now have 512 of it. but that doesnt mean it is now balanced for next redering solution.

This doesn't make the slightest sense.

 

[SIGSEGV]

Completely out of curiosity, what heatsink did you use that only cost $40? The Accelero Xtreme III that everyone is recommending starts at $75.

50 bucks is enough to cool 290x

 

[Aerpoweron]

Does the 780 TI support DirectX 11.2?

Nope - just found out:
GTX 780 Ti GPU Engine Specs:
2880CUDA Cores
875Base Clock (MHz)
928Boost Clock (MHz)
210Texture Fill Rate (GigaTexels/sec)
GTX 780 Ti Memory Specs:
7.0 GbpsMemory Clock
3072 MBStandard Memory Config
GDDR5Memory Interface
384-bitMemory Interface Width
336Memory Bandwidth (GB/sec)
GTX 780 Ti Support:
4.3OpenGL
PCI Express 3.0 Bus Support
YesCertified for Windows 7, Windows 8, Windows Vista or Windows XP
GPU Boost 2.0, 3D Vision, CUDA, DirectX 11, PhysX, TXAA, Adaptive VSync, FXAA, NVIDIA Surround, SLI-readySupported Technologies
Yes3D Vision Ready
Microsoft DirectX 11.1 API

Is it the real DX11.1 this time, or just 11.0 like on the other 7xx cards?

Nvidia is still confusing me with their Direct X labeling.