[TechSpot and Tom's] Crysis 3 GPU and CPU benchmarks

[2is]

Screwing around with CPU time cycles for a game, or for that matter, any application is beyond asking for trouble.

I mean, I'm no hardware expert, but the engineers at Intel are. I trust them more than some muppet in a basement hacking hex code to change the behavior of important and relatively expensive hardware.

I too thought of using that until I read more about it. It could very well work without issue for most. Yet, I am not willing to take that risk just for a game IMO.

I take it you don't overclock either?

 

[omeds]

Just came across this. By any chance, any of you tried this ?
Crysis 3 - 30% Framerate and Performance Boost, and FPS Cap Removal

His system is borked, most likely from tweaking. He has 10.001ms by default, I have 0.97ms by default and the tool doesnt increase fps at all for me, even at low res with minimal settings to remove GPU load. He only sees an increase because the tool is setting his borked system back to near default levels.

 

[guskline]

http://www.techspot.com/review/642-crysis-3-performance/

Crysis 3 is out and is the system-killer it was designed to be.

I've seen lots of posts on the forum from people complaining that their new rigs can't play the game at max. Kind of ironic because the complaint leveled at most other games is that new rigs CAN play them at max.

Anyway, looks like a graphics tour-de-force and a benchmark we'll all be using for years to come.

Two observations:
- AMD needs some updated drivers
- the game is very sensitive to CPU threads and speeds. OC'ing brings big gains at medium settings with both Ivy Bridge and Vishera.

I'm having fun with it on my FX 8350 rig @4.6 and a GTX680. I'm using an Achieve Shimian 27" 2560x1440 ("poor man's Dell") running at 2560x1440. I opted for FXAA and High settings to have playable frame rates. No doubt about it. Crysis 3 puts a stress on the system. I'm working my way through the campaign and the FX8350 is holding up well.

 

[Termie]

Tom's has conducted some new Crysis 3 benchmarks, testing low-end all the way through Titan, as well as a number of CPUs. It's also using it's own latency testing, which may provide a bit of additional insight into VGA performance. My personal opinion is that Tom's is a bit confused about what type of variation actually negatively impacts performance.

Either way, here's the link: http://www.tomshardware.com/reviews/crysis-3-performance-benchmark-gaming,3451.html

 

[notty22]

scaling at 1080p not great.

 

[Attic]

After playing for a bit i've had to settle with Low Textures, NoAA, Medium system spec at 1080P to maintain 35-50fps. Also noted game is very CPU intensive.

That being said even at those near bear minimum, quality looks very very good. I get a slide show at Very High settings,.... everything looks a bit more wet... more alive at the higher detail.

A game I will come back to play once I upgrade into 20nm GPU's.

 

[Carfax83]

A game I will come back to play once I upgrade into 20nm GPU's.

Same here, except I haven't bought the game yet. I don't want to play Crysis 3 until I can come close to maxing it out comfortably @ 2560x1440. I was going to buy it and upgrade to 7970 Ghz CF or 680 SLi, but after much thinking, I decided to wait for the next generation.

Crossfire still appears to be quite buggy, and the 680 doesn't really impress me that much in terms of a performance gain over the 580. Titan is too expensive as well..

So I'll go play Tomb Raider and Bioshock Infinite when it becomes available later this month instead :biggrin:

 

[crashtech]

Bump, looking at some of these results while deciding what video card to buy has me hesitating. Hope no one minds me bringing this to the top to ask why the 670 (my second choice) seems to be holding its own so well against the 7970, which most people seem to agree is superior.

 

[ShadowOfMyself]

Game seems to favor Nvidia cards heavily... Hoping AMD will make some "magical" drivers that their performance on par with other games

Anyway, as I thought, there is barely any difference from low to very high settings... Shadows get more defined and the shading option is the only one that makes a big difference, since it seems to control tesselation

That said the game looks better on low settings than most games out there, and performs well too, so good job on them optimizing it

 

[Will Robinson]

Game seems to favor Nvidia cards heavily... Hoping AMD will make some "magical" drivers that their performance on par with other games

Anyway, as I thought, there is barely any difference from low to very high settings... Shadows get more defined and the shading option is the only one that makes a big difference, since it seems to control tesselation

That said the game looks better on low settings than most games out there, and performs well too, so good job on them optimizing it

AMD must have only gotten the code a week ago......'splains everything...

 

[tviceman]

AMD must have only gotten the code a week ago......'splains everything...

They paid for the bundle months ago but only got the code a week ago. Gotcha. Gaming Evolved at it's finest.

 

[blackened23]

Ehh? They seem about even (680 vs 7970) over at hexus.net with the latest drivers for both cards, IIRC didn't AMD release an updated driver with improvements shortly after release? Benchmarks from Feb probably do not take that into account..

http://hexus.net/tech/reviews/graphics/52329-club3d-radeon-hd-7990-6gb/?page=5

and HardOCP dated on the 12th: These were JUST done with the latest drivers for both sides:

Multi GPU scaling:

Also, Xbit labs:

Pretty much a dead heat and same speed between 7970ghz and 680. These were done at the absolute highest quality setting with FXAA as far as I can tell.

 

[Xarick]

He was talking about the 670 and the 7950 boost,
But again.. the 670 is the faster card.. the 7950 is faster when Overclocked.

 

[Gikaseixas]

Yep 670 > 7950 and 680 < 7970
This is one of the closest battles in last few years

 

[BallaTheFeared]

Yep 670 > 7950 and 680 < 7970
This is one of the closest battles in last few years

You forgot <<<<< Titan

 

[Will Robinson]

Titan gets to play with other cards in its price range ne?

GTX670 and HD7950 are pretty evenly matched per se but 680 V 7970GHz is a decent win for the Radeon.:thumbsup:

 

[3DVagabond]

Bump, looking at some of these results while deciding what video card to buy has me hesitating. Hope no one minds me bringing this to the top to ask why the 670 (my second choice) seems to be holding its own so well against the 7970, which most people seem to agree is superior.

It's one game. Taken across a range of games, the 7970 is the faster card. The additional gig of ram is a plus as well.

 

[Ares202]

It's one game. Taken across a range of games, the 7970 is the faster card. The additional gig of ram is a plus as well.

Resolution is a big factor also, 1920x1200 and under the Nvidia cards seem to win most of the time, resolutions higher than this the Radeon cards have an advantage.

I went with Nvidia because the vendors are simply better, AMD needs to do more to get the better vendors on its side.

 

[3DVagabond]

Resolution is a big factor also, 1920x1200 and under the Nvidia cards seem to win most of the time, resolutions higher than this the Radeon cards have an advantage.

I went with Nvidia because the vendors are simply better, AMD needs to do more to get the better vendors on its side.

You need to read recent reviews. The 670 is simply not faster than the 7970 most of the time. Release reviews were, but not anymore.

 

[crashtech]

Ehh? They seem about even (680 vs 7970) over at hexus.net with the latest drivers for both cards, IIRC didn't AMD release an updated driver with improvements shortly after release? Benchmarks from Feb probably do not take that into account..

http://hexus.net/tech/reviews/graphics/52329-club3d-radeon-hd-7990-6gb/?page=5

and HardOCP dated on the 12th: These were JUST done with the latest drivers for both sides:

Multi GPU scaling:

Also, Xbit labs:

Pretty much a dead heat and same speed between 7970ghz and 680. These were done at the absolute highest quality setting with FXAA as far as I can tell.

Thanks for the updated info. These latest results seem to reflect the capability of the hardware more realistically. Now my choice seems to be between the 680 and the 7970 GHz.

 

[Ares202]

You need to read recent reviews. The 670 is simply not faster than the 7970 most of the time. Release reviews were, but not anymore.

Where in my post did i compare the 670 and 7970?

I was trying to illistrate that the additional vram and memory bandwidth only really comes into play after 1920x1200

 

[exar333]

scaling at 1080p not great.

The CPU scaling is VERY telling IMO

-Min FPS almost doubles from 2C-> 4C while average FPS increases about 25%
-Min FPS goes up about 10% from 4C -> 6C but avg FPS increases ~25%
-FX is abysmal; barely faster than i3 2C, although a MP benchmark might show more difference. Based on the config shown in the below graph, ANY AMD CPU is essentially unplayable at that setting...

http://www.tomshardware.com/reviews/crysis-3-performance-benchmark-gaming,3451-8.html

 

[crashtech]

It looked like the benchmark run had an anomaly near the end that brought the minimum frame rate way down on every configuration that was run. I wish they would have run some other runs to verify if that kind of slowdown was really representative of the game. It even states in the review that there wasn't even much happening on the screen during the anomalous segment, so it remains a mystery, and may be a glitch in the benchmark.

 

[Elfear]

Resolution is a big factor also, 1920x1200 and under the Nvidia cards seem to win most of the time, resolutions higher than this the Radeon cards have an advantage.

I went with Nvidia because the vendors are simply better, AMD needs to do more to get the better vendors on its side.

Actually the 7970Ghz does pretty well at 1080p. Xbit's review came out yesterday.

 

[3DVagabond]

Bump, looking at some of these results while deciding what video card to buy has me hesitating. Hope no one minds me bringing this to the top to ask why the 670 (my second choice) seems to be holding its own so well against the 7970, which most people seem to agree is superior.

Where in my post did i compare the 670 and 7970?

I was trying to illistrate that the additional vram and memory bandwidth only really comes into play after 1920x1200

Here's the original post I was responding to. He was comparing the 7970 with the 670. That's where I posted it was only one game, and then you started responding to me. It was the 7970 vs. 670 from the start.