Review: Far Cry Primal PC graphics card performance benchmarks

It's well known that AMD cannot optimize for specific games until the game has been released if they utilize GameWorks (FC4).

FC4 was one of the biggest messes of GW to date at release. It's decently stable now. I'm amused to hell and back to find that it's not being used in Primal.

Not to mention optimizations are specific for Kepler, as they are for Maxwell. nVIDIA doesn't even take care of Kepler anymore. This was obvious when they released a "bugged" driver, that they admitted, at the release of W3 that was degrading performance.

I'm pretty sure the tinfoil hate theories were put to rest last year. Kepler isn't getting special care with optimizations like Maxwell is. Meanwhile, GCN gets constant updates.

If there was heavy tessellation use in Primal, the Kepler cards would be even slower.

In the Primal graph, the 980 is only 12% faster than the 780 Ti in avg fps, and 15% faster with minimum fps. /fin foil hat

Nothing to see here.

 

Nope. AMD has basically showed that it doesn't even care to optimize some games. Look at Project Cars. Devs gave AMD access to the game and AMD made no effort to contact the devs. Game came out and they cried the blues about Gameworks and source code. Same thing with Witcher 3. AMD claimed it couldn't optimize the game because of Gameworks being closed source. A month later they released a driver that fixed the performance problems, without ever getting access to the source. Sorry but that argument doesn't hold water.

Not to mention that AMD does the same thing. Remember when TressFX came out and Tomb Raider was a buggy mess with it enabled on Nvidia hardware? Yeah, same crap.

So you're essentially saying any kind of driver regression = deliberate sabotage by the company? Ok bro. And lets keep pretending that in the face of these accusations that the company just continues to downgrade it's products, including now even Maxwell, which it's still currently selling. That makes a whole ton of sense.

Just kidding, it's the dumbest argument ever.

 

Nope to what, that they can't optimize after release? Because they can, it's in the disclosure by nGreedia.

If they don't optimize for some games then who knows why?

Project Cars is not only Gameworks based, but heavily PhysX driven, no pun intended. So now, AMD with its CPU overhead issues, now has to deal with even more CPU issues.

http://hardforum.com/threads/projec...ormance-issues.1861353/page-2#post-1041593386

The day PhysX works with AMD is the dead cows fly and you become unbiased.

The developers were happy to make a patch for nVIDIA, which I guess helped. I finished Tomb Raider with 680 SLI. You know, because it's open source, like AMD intended?

https://community.amd.com/community/gaming/blog/2015/05/12/tressfx-hair-cross-platform-and-v20

Uhhh, no. How on earth did you deduce that?

Kepler, as speculated by many, isn't getting specifically cared for optimizations that Maxwell is getting.

It's not an argument. It's either true or false, it's not a matter of opinion.

 

Oh yes, that game was amazing (except being kicked by punkbuster for using the official HDR patch on multi-player matches -.- )

 

Must be intentionally downgrading.

You mean like how CD Project Red created patches for AMD and how AMD users were also able to complete Witcher 3? Same with Farcry 4? So basically the same thing?

Actually it is an argument because you haven't proved it. You can show evidence of Kepler getting slower all day, none of that is linked to lack of optimization.

 

They weren't for AMD users directly. Did you not see the amount of Kepler owners crying? Single card Maxwell users were even happy for it. Do you see a tessellation setting in the nVIDIA drivers that literally, zero people on this Earth have seen?



You seem to be confusing an optional setting with direct code optimizing. Shocking.

Yeah, totally open source like TressFX. Some nasty crappy algorithm.


Actually it is an argument because you haven't proved it. You can show evidence of Kepler getting slower all day, none of that is linked to lack of optimization.

 

A Fury X performs slower than a 780Ti in TessMark. There is no software optimization for a lack of hardware. Isn't that what you said about Async?

 

You can't directly see what tessellation limit driver has for given game. (And I think that since benches were done at 16.2 which did not have Primal profile, there in no limit = 64x.)

16.2.1 added profile. Anyway it can be tested by trial and error. Bench at default profile, then overriding it to allow maximum of 64x... look for difference in performance, retry at 32x limit, 16x, ...

Easy to find how it is set up. Anyway, at 1080p 8x is usually more than enough and difference from x64 is not visible in motion.
At 4k 16x tessellation limit behaves same way.

 

Yeah, I guess my question is, does AMD force the profile to a lower amount and is Nvidia doing the same in their driver?

Like is the game, or any game for that matter, being forced to 8x on AMD yet is still running at 64x on Nvidia cards? I would imagine that AMD lets the developer set the level and goes with it. And I agree, in general it doesn't make much of a difference after a certain point. There are a couple of exceptions. God Rays in FO4 for example, get weird around dense objects (like fences) when the factor is turned lower.

Nvidia should really add a slider to their CP. Actually they should just scrap the entire CP and move everything to experience. Their control panel in general is just old and terrible.

 

I am sorry to say, but Tessmark 0.3.0 is fundamentally flawed.
I get from it following fps (Default settings X16 - Normal):
Heating up at start, Benchmark window is not focused, and is at least partially covered by its launcher (one pixel in corner or more):1560~1570fps
Heating up at start, window is focused: 1440~1480fps
Heating up at start, bench window is not focused and not covered: 1370~1380fps
Benchmark window is not focused, and is at least partially covered by its launcher (one pixel in corner or more): 1310~1320fps
Benching while window is focused: 1200~1220fps
Other window is focused, but bench is still fully visible: 1120~1130fps

And this is with GPU running 1050MHz. And now guess ho many fps I get when I downclock GPU to 800MHz.
Benching while window is focused @800MHz: same 1200~1220fps as above
(And even at that moment GPU utilization is like 60~70%.)

Tessmark 0.3.0 simply has some kind of limitation which collides with either desktop composition or AMD driver as its fps and therefore score is not directly affected by GPU clock and is actually affected by mere window focus and fact that at least one pixel is covered by other window.

And here is something to laugh at:

Spoiler

Btw if what is here is to be believed "These TessMark 0.3.0 results are also valid with GpuTest 0.6.0.".
I do score there at settings matching tessmarks: 104966 points with average 1749fps
www.geeks3d.com/20131108/geforce-gtx-780-ti-the-king-of-tessellation/

 

AFAIK the default tessellation mode in CCC is set to "AMD Optimized", effectively limiting max tessellation level on a per application basis, that is as long as a preset was made for the application.

you can bet witcher 3 later received the optimized tessellation level treatment at the very least. however, I do not recall AMD ever specifying the exact tessellation level set for a given title.
I also wonder if they can dynamically limit the level in-game, based on tessellation amount, GPU load, etc.. I would appreciate some effort in that direction, rather than pick a number and call it a day.
(not to confuse with distance driven LOD / dynamic tessellation implementations which are at the hands of devs. I don't think the driver can or does manipulate any of that, beside capping the factor)

as for Nvidia, they could theoretically do the same, but never did they communicate nor expose such 'driver feature'. if anything, they kind of promote high tess levels through their own special effects and guides.
acknowledging an advantage in tess performance I'd say they most likely do not limit it at the driver level. with that said they do implement dynamic tessellation optimizations in their GameWorks libraries.

 

Multi-GPU: the rest of the story...

I play in Nvidia Surround (5892x1080) and to make it playable I use three 980ti cards. Problem is I'm getting 10 fps less than 980ti SLI on 4K benchmarks listed here (which is more pixels than my surround setup) on lower graphics settings. This game doesn't like more than 2 cards... No big loss, it's merely a stop-gap until The Division launches Tuesday ;-) Now there's a well optimized game! Max settings and 80-100fps in surround....at least the Beta seemed to be well optimized. Hope the full release is the same...can't wait!

 

Well, If I would do it. I would use method which limits tessellation based on how many pixels polygon takes on screen.
Above mentioned Tessmark in x32/x32 mode calculates and renders 2.764 million polygons and there I maintain around 680fps. It is at default 1280x720.
Fullscreen 1920x1080 improves that fps to around 700.

And that is still 2.764M polygons on 2.074M pixels. Kind of waste.
tessellation x64/x64 pushes 11.059M polygons, fps goes down to 210 on 1920x1080.
While tessellation reduces fps, it does not reduce it enough by itself to cause sub optimal gameplay even while having 5 polygons per pixel or more.
Maybe in Witcher 3 it was in combination with "special" AA applied to Tessellated hair.
Because in end of the day AA actually takes into account polygon edges and maybe AMD's implementation does not like much to calculate state of pixels where there are dozens of small polygons "INSIDE". As usual AA deals with edges of polygons, not entire polygons on sub pixel sizes.

Edit: Unlikely to be AA, as in x64 mode I have with Hairworks around 61fps and changing HairWorksAALevel=4 to 0 makes like 1.5fps difference only. While having tessellation at x8 moves fps to 89.

 

This game rules just about to finish it already finished Batari going for Udam homeland now. Totally cleaned the map of everything except the collectibles(since map you cant enable them and see the missions at the same time) so I'm gonna sweep those on the SaberTooth after I beat the game. I thought it would be lesser in content being only 16gb but I guess 10,000bc has more repetitive textures(looks like way more graphics but only half the install size). Yeah game is repetitive they nailed the looks of it gameplay could have been improved. Has a kind of brutal feel to it which could have been taken to extreme heights to where the game really felt like 10,000 years ago. If they would have matched the powerup progression made it less haokey made it less of a far cry(a ubisoft farcry)(at least this one isn't like mtv or communist revolution like the last 2).

Sure is funny to try to push liberalism and multiculturalism in a game set in 10,000BC and is largely about revenge and genocide...... Really the game itself as far as the storyline makes no sense how come they are black African tribesman in prehistoric Europe(didn't happen). I can see this offending minorities too because the hero is white and so is his best buddy but the rest of the tribe is black(also not very realistic but a typical frenchie mtv tactic) like they need white men to bring them everything and save them from being destroyed by the other tribes(the multiculturalism in this game actually comes off kinda racist if you really look at it closely). How come you go around killing all the stronger tribes so your lazy ghetto(man just look at the hairstyles in the game before you go ape over this assessment) tribe can take over(cause they are good people and liberals? oh yeah you got a shaman that's always feeding u drugs so your a cool guy aka a liberal drug addict just like far cry 4, far cry 3 and now gta v all have the kill everything while your tripped out and high missions, I enjoy those to be honest but its just pushing liberalism). They make a southern us hick the dumbass comic relief of the game and the tribe(that uses mtv rap music terminology mixed into his prehistoric babble bootyqua I picked out some other things in there but don't remember them all)(whos fighting all ur middle east wars frenchies?(southern us hicks) oh that's right ur just gonna hump them all to death Charlie Hebdo style) the same sounds at times like the prehistoric babble is French and the game seems to be set in French setting with the alps in the background and the French countryside. Too bad my French isn't better otherwise I could draw more parallels between the ingame languages and French which I'm sure exist. Yeap lead creative producer is Frenchman so ubi stole far cry from cevat and the turks(crytek has since become mostly turk coders and designers) and have turned it into some French controlled production and they tried to hide it from an American audience but anyone who has any experience with French media will see the parallels in this game and the prehistoric babble used in the game. Really its a French game that mascarades as an American one but if you are French you will see its a French game.

If the progression was perfectly tuned and harder this would be a formidable game when you run around at first it seems incredibly hard and those sabertooths pretty much if one sees you your dead or any cat for that matter. Also the caves give it a great prehistoric feel but the mazes are too simple maybe one had a good trick in it in the very north of the map but the rest were very easy. The game looks great but the problem with it is its way too easy I died a few times in the game and it was mostly from not understanding the interface. LOL TOOK ME MOST OF THE GAME TO REALIZE THAT TO CHANGE THE WEAPONS IN THE WEAPON WHEEL YOU HAVE TO USE THE MOUSE WHEEL........ Also why does it tell you when you get the multiple item power ups that resource usage is the same when in actuality you multiply your weapons per resource by the multiple it says in the skill tree. So really resource usage is divided by the multiple whoever made the interface on this game is not very bright(or its translated poorly from French or by a Frenchman doing the English interface). Feeding the Sabertooth to ride it is a pain and petting and some of the feeding animations take too long. When you get all sting and berserk and fire bombs you can wipe out an entire outpost without even having to fight making the battles at the end of the game extremely easy. The graphics and mechanics are amazing and they really improved on the dunia 2 engines speed and smoothness from the last 2 farcrys but the lod was upped so much your going to be in the 40-60fps range(so even tho the engine is faster smoother improved overall the game is slow and laggy unless your on the latest hardware/ and who wants to upgrade for this game.... no one) unless ur on dual 980tis or there abouts and in the northern mammoth hunt area your going to be in the 30s most of that map.

The loot system is kind of a double edged sword. For a clean sweeping nut like me I like to loot everything but because of the basic loot and crafting really it is unnecessary really you can play the whole game with just stash loot which there is a ton of the rare loot in the caves or if you get the skill upgrade u will get rare loot just looting the countryside. Obviously if you loot everything you will just be able to craft all the weapons sooner and it will make the game easier, but later on in the game if you get tired of looting u can ignore it all together fast travel or hit up the local camp stash and replenish yourself for the next outpost or battle(or if you use the skill trees for double weapon creation you will stretch your resource usage much further which is the opposite of what the skill tree tells you). Also the loot limits of the pack(which have limits per item but in reality would add up to thousands of pounds that no real person could realistically carry) are unrealistic but kind of nice because it makes looking and accumulating for building the various huts in the map easier to understand but different from other games loot structures in other open world games(you wont be transferring stashes like Stalker Shadow of Chernobyl).

Pros: Great mechanics, Great Graphics, Smooth engine much less glitch than before, huge open world with high level of detail, great caveman feel, riding fast beasts around the map while being able to aim down sight and kill things while riding as well as search them except to take the missions where u have to get off to talk to the guy to go have the mission added to your map and run over there to actually go do it(these missions became a pain)

Cons: Frenchie liberal brainwash, makes fun of white americans aggrandizes everyone else, ridiculous non realistic multiculturalism wash, for some reason everyone thinks prehistoric tribes in Europe were all black....., same gameplay mechanic as farcry 4 pretty much, WAY TOO EASY, progression not tuned for good survivalist feel turns it into pretty much a completionist far cry, some of the gameplay modes are long and boring, feels like a childs game

Overall: recommend because what far cry is now is what gamers dreamed about 20 years ago, the problem is ubisoft and their liberal brainwash agenda not the game coders or artists they are doing their job right, if ubisoft didn't have to turn out mtv impersonation crap(remember the far cry 3 commercial?) they could be making the best games in the industry right now

OH YEAH HAS DENUVO WONT BE CRACKED FOR YEARS SO STOP WAITING MIGHT AS WELL JUST PAY FOR IT NOW INSTEAD OF WAITING FOR 2 OR 3 YEARS TO GET IT AT 15 BUCKS WHEN U CAN GET IT FOR 32 BUCKS NOW.

GAME RUNS FAR BETTER ON 361.91 DRIVERS UNLESS YOU HAVE 980TIS ANYTHING LESSER .91 IS A FAR SUPERIOR DRIVER TO 362.00 364.47 OR 364.51(ALL THESE DRIVERS SUCK UNLESS YOUR ON A 980TI OR 980) I GET 48 FPS IN OPEN SPACES WITH .91 AND MSI GAMING APP ON. WITH ANY OF THE OTHER DRIVERS I GET EVEN LESS AND YES .91 HAS THE SLI PROFILE. OH YEAH INSTALL PROCESS LASSO SET GAME PROCESS, EXCLUDE PROBALANCE RESTRAINT, AND SET AFFINITY TO NON HYPER THREADED CORES THIS WAY YOU DONT HAVE TO DISABLE HYPERTHREADING IN THE BIOS AND YOU CAN USE HYPERTHREADING IN THE OS WHEN YOU NEED IT AND JUST TURN IT OFF FOR THE SPECIFIC GAMING PROCESS.

Wow played with voltages in bios and it helped this game out tremendously. Kinda mad now I played it laggy now I'm getting sometimes still 40s in low lying open spaces but now on mountains im getting 50-60 and in the nortern snow areas im getting a solid 46+- FPS. Game is real fluid and fast now just like a console game. Kinda mad now I already played it before it was this quick. Oh well. Make sure you self tune every voltage in your bios cause they dont self configure as well as they used to.

 

Can someone with an overclocked 970 run the benchmark at the same settings as the guru3d review 1080p or 1440p and post their results and clocks?