[techpowerup] AMD R9 290X vs nvidia Titan benchmarks

[Grooveriding]

http://www.techpowerup.com/191768/radeon-r9-290x-clock-speeds-surface-benchmarked.html

 

[VulgarDisplay]

I'm still sticking with what I said in another thread. It will be entirely game dependent on which is faster. The caveat being that the 290x will be faster on average in AMD games than the Titan is in nvidia favored games.

At least until Mantle comes out and it "ridicules" the Titan.

 

[3DVagabond]

From the article:

Over all, at 800 MHz (core) and 4.50 GHz (memory), it's advantage-AMD, looking at these graphs. Then again, we must remind you that this is 5760 x 1080 we're talking about. Many Thanks to Grant Kim.

If these clocks are real there's going to be just a bit of O/C headroom left.

I don't understand why AMD did this? Unless they are going to be higher on the retail cards and were trying to blindside nVidia with "leaked benches".

 

[Udgnim]

From the article:


If these clocks are real there's going to be just a bit of O/C headroom left.

I don't understand why AMD did this? Unless they are going to be higher on the retail cards and were trying to blindside nVidia with "leaked benches".

if 300W TDP is true, won't OCing be wattage limited?

 

[hyrule4927]

As TechPowerUp mentioned, it is a little odd that the memory bandwidth is below what AMD said last week. Also, what the heck is up with the framerate in Rage?

 

[3DVagabond]

if 300W TDP is true, won't OCing be wattage limited?

What 300w TDP? I missed where TDP was leaked/released.

 

[VulgarDisplay]

if 300W TDP is true, won't OCing be wattage limited?

The only leak we've seen so far that shows power draw has it as less than the titan. Everyone is making this 300w TDP number up based on the connectors on the card I'm pretty sure.

 

[Grooveriding]

if 300W TDP is true, won't OCing be wattage limited?

Even if that number is accurate it won't matter to overclockers. Just like we have overclocked GTX 780s and Titans gobbling up 400W, the story will be the same, you just need to be able to keep it cool enough.

 

[Gikaseixas]

800mhz :twisted:
Granted there's some overcloking capability this is a monster card.
I want some 512-bit goodness for my i7 4770k build

 

[AnandThenMan]

Sounds like a pre-production/engineering sample was tested.

 

[BoFox]

If the rumored specs are correct:

2048sp @ ~1035 MHz for 7970GHz = ~4.24 TFLOPs

2760sp @ 800MHz for R9 290X = ~4.42 TFLOPs

That's hardly any increase. It doesn't add up. There must be something else for it, unless the ROPs were badly bottlenecking the 7970GHz in 5760 resolutions. I think there must have been an hidden boost clock that was not detectable by anything yet.

Or there was even more shader processors, perhaps as many as 3000+...

 

[hawtdawg]

I doubt this is going to make 780 prices drop very much.

 

[Saylick]

If the rumored specs are correct:

2048sp @ ~1035 MHz for 7970GHz = ~4.24 TFLOPs

2760sp @ 800MHz for R9 290X = ~4.42 TFLOPs

That's hardly any increase. It doesn't add up. There must be something else for it, unless the ROPs were badly bottlenecking the 7970GHz in 5760 resolutions. I think there must have been an hidden boost clock that was not detectable by anything yet.

Or there was even more shader processors, perhaps as many as 3000+...

Assuming the 2816 SP information is true, and given the >5 TFLOPS of crunching power, we should be looking at a minimum average clock speed of 887.8 Mhz.

There is also something fishy about the 288 GB/s of bandwidth as well, considering the reported value was > 300 GB/s.

I'm thinking there's some kind of boost mode as well. Either way, 800 MHz is on the low side. If 900 Mhz was the stock clock, I'd imagine you could easily eek out another 200 MHz.

 

[HurleyBird]

Given that they would directly contradict the GPU14 info that was just given out, these probably aren't final clocks. It's also possible that the clocks just aren't being reported correctly, because the performance seems out of this world for 800MHz core and 1125 MHz memory.

On the other hand, maybe it's just another AWESUM switch and AMD reported those figures at GPU14?

 

[StrangerGuy]

As TechPowerUp mentioned, it is a little odd that the memory bandwidth is below what AMD said last week. Also, what the heck is up with the framerate in Rage?

LOL Rage. Like who cares how it performs in this bad game.

 

[x3sphere]

As TechPowerUp mentioned, it is a little odd that the memory bandwidth is below what AMD said last week. Also, what the heck is up with the framerate in Rage?

Rage is OpenGL. They had problems with it at launch too.

 

[RussianSensation]

800mhz :twisted:
Granted there's some overcloking capability this is a monster card.
I want some 512-bit goodness for my i7 4770k build

That sounds hard to believe since AMD said > 5Tflops of stock performance. That implies a GPU clock speed of at least 890mhz. If true, this is going to be be overclocker's dream with unlocked voltage! It takes roughly a 1.05-1.1Ghz 780 to match a Titan iirc. If this card is matching the Titan at only 900mhz, assuming these can clock to 1.3Ghz, it should be faster than the 780 overall.

 

[blackwhiskers]

Rage is OpenGL. They had problems with it at launch too.

AMD still has problems with OpenGL in other games, too. This is one area they could really improve their drivers. These benchmarks just show that at least preliminary nothing has changed on this front :/

 

[Silverforce11]

Yup im excited about the low stock clocks, massive OC as a selling point again, well done AMD, leaving enthusiasts to unlock the beast without locking down voltages. It could well be a stock 250W part like Titan, but once people OC, power use skyrockets as does performance.

 

[Imouto]

These are fake as hell, notice how the average frame rate for the Titan in the first graph is lower than the min figure.

 

[Techhog]

These are fake as hell, notice how the average frame rate for the Titan in the first graph is lower than the min figure.

Yep. Worst fake ever.

 

[LegSWAT]

LOL Rage. Like who cares how it performs in this bad game.

Rage was the first game ever that made me delete the intro video with bethesda logo from the base folder to actually get the game running. Either way, I didn't want to watch that logo anyway and I finally got that thing running
On my Win7/64 I didn't even bother reinstalling the outdated version of MS dotnet fx 3.5 framework to give that game another chance. What a piece of garbage!

 

[caswow]

This is fake. dont need to discuss further.

 

[King Mustard]

If true (I doubt these benchmarks are genuine), it seems the performance is similar, but the AMD card is mostly likely going to be much cheaper.

To be fair, it's not too fair to compare a NVIDIA card that came out in February with an AMD card that's not even out yet (it's now October).

 

[blackened23]

Well, these results pretty much line up with the leaked results at Chiphell as well - The dudes at chiphell are from China/TW so a few of those guys have access to hardware that we don't, with no concerns of NDA. (*all PC hardware is mass produced over there, so they have much more reliable info than we do*) The leaks from chiphell with the 7970, GTX 680, and GTX Titan were all accurate in terms of benchmarks. Still, take them with a grain of salt - there are a lot of variables still not reported. Such as: will single screen benchmarks align with eyefinity results? Will non AA benchmarks be as strong as AA benchmarks? Again, a lot of unknown variables. We'll just need to wait until the 15th.

So i'm not putting complete faith in leaks. Even if they are real, nvidia will definitely have it's share of benchmark wins - some games just favor NV's architecture, notably Ubisoft and blizzard games. So i'm sure things will "even out", so to speak. Yet, despite that, if these benchmarks are indeed real (and TPU did note that the benchmarker validated everything via GPU-Z upload) it is highly impressive - trading blows with the Titan for 600$. Not too shabby.