[H] Battlefield 4 Video Card Performance and IQ Review

[blackened23]

Did anyone complain about BF3 multiplayer benchmarks at H?

http://www.hardocp.com/article/2011...ltiplayer_performance_iq_review/#.UoqazcSkreM

 

[3DVagabond]

That is actually a good point as to why such benchmarks should not be done imho. They are unreliable. And if so, what is the point to even doing them?

Because companies cheat with the canned benchmarks. Depending on which company backed the game they will make sure that the benchmark shows their products strengths over the other brand.

 

[Teizo]

Did anyone complain about BF3 multiplayer benchmarks at H?

http://www.hardocp.com/article/2011...ltiplayer_performance_iq_review/#.UoqazcSkreM

Looks like a clean sweep for Nvidia. Good review!

Because companies cheat with the canned benchmarks. Depending on which company backed the game they will make sure that the benchmark shows their products strengths over the other brand.

Yeah, that is true. However, everyone knows there is really no way for a stock 760 to beat a stock 770 though unless something funky is going on. Which is why I think it needs to be prefaced this game is still in beta and performance numbers are not reliable.

 

[3DVagabond]

Yeah, that is true. However, everyone knows there is really no way for a stock 760 to beat a stock 770 though unless something funky is going on. Which is why I think it needs to be prefaced this game is still in beta and performance numbers are not reliable.

No argument there, about the 760 vs. 770. Although Brent did say they used a retail version of the game, not a beta.

 

[Slomo4shO]

Yeah, that is true. However, everyone knows there is really no way for a stock 760 to beat a stock 770 though unless something funky is going on. Which is why I think it needs to be prefaced this game is still in beta and performance numbers are not reliable.

That is the full game review, not the beta... Also, was the paragraph detailing the issue below the graph or the paragraphs in the conclusion not sufficient in explaining the anomaly for you?

 

[3DVagabond]

That is the full game review, not the beta... Also, was the paragraph detailing the issue below the graph or the paragraphs in the conclusion not sufficient in explaining the anomaly for you?

I think the point is that the anomaly shouldn't exist. It raises the question about possible other anomalies that might skew the other results as well. I'm actually surprised they left the 760 score like that and didn't try and get a better representative score.

 

[Teizo]

I know they used a retail copy. It is my assertion that the EA released the game still in Beta because they 'had' to release it before Call of Duty. Once they patch it and get it fixed then the game will be what it was supposed to be upon launch and performance numbers can be seen as more accurate and worthwhile.

 

[blackened23]

Here's the argument I think they would make: HardOCP doesn't review games. They review GPUs and other assorted PC hardware. They've always tested AAA games near launch time, always, always without exception.

BF4 isn't in such a state that it can't be benchmarked. And waiting a month for both parties to get their crap together is a slippery slope to say the least - certainly there have been NV centric games that were benchmarked a week after launch without an AMD driver. It's up to AMD and nvidia to have their game ready drivers out. But the readership, IMHO, likes getting performance data for a game near the release of the game in question. Again, they're not game reviewers. PC hardware reviewers. Big difference. They're not going to wait for game patches, if it's able to benchmarked they will do it. Waiting for AMD or nvidia to release their own game drivers is also a slippery slope. These games just need to perform at their best near the launch date.

Are you playing BF4? Just curious. I mean, i've heard a lot of stories about BF4 being buggy, but really not any more-so than BF3 (I put a TON of hours into BF3..). I had a crash or two or three or four in the first couple of days, but when I disabled the afterburner overlay they ceased. I dunno. Maybe i'm lucky.

 

[Skurge]

I can tell you these benches are useful. If it werent for H testing MP, who would know 6 series AMD cards generally struggled in the game even though the FPS didn't show it?

Same with the stuttering that Nvidia cards had in beta is still present in the retail version even if the numbers don't show it.

I welcome [H]'s data as they play through the game. Some other reviewers won't be able to tell you why a 760 is faster than a 770 because they wouldn't be playing the game. Instead just running time demo's.

 

[Skurge]

Here's the argument I think they would make: HardOCP doesn't review games. They review GPUs and other assorted PC hardware. They've always tested AAA games near launch time, always, always without exception.

BF4 isn't in such a state that it can't be benchmarked. And waiting a month for both parties to get their crap together is a slippery slope to say the least - certainly there have been NV centric games that were benchmarked a week after launch without an AMD driver. It's up to AMD and nvidia to have their game ready drivers out. But the readership, IMHO, likes getting performance data for a game near the release of the game in question. Again, they're not game reviewers. PC hardware reviewers. Big difference. They're not going to wait for game patches, if it's able to benchmarked they will do it. Waiting for AMD or nvidia to release their own game drivers is also a slippery slope. These games just need to perform at their best near the launch date.

Are you playing BF4? Just curious. I mean, i've heard a lot of stories about BF4 being buggy, but really not any more-so than BF3 (I put a TON of hours into BF3..). I had a crash or two or three or four in the first couple of days, but when I disabled the afterburner overlay they ceased. I dunno. Maybe i'm lucky.

I agree. If these numbers aren't shown. It won't push AMD or Nv to fix any performance issues. Like Dragon Age 2 for Nvidia or BF3 for AMD and a few other cases. It's better for all us.

 

[blackened23]

I think it's fine, yeah. I also don't think the 290X winning against the Titan/780 makes it a better card, either. So, I don't really see why some people get bent out of shape. One game where the so-called winner doesn't earn the "better card" title. That's how I see it anyway - but I like having the performance data nonetheless.

I'm sure Brent and company will do another follow-up with the 780ti, that's where things will get interesting if Nvidia has a game ready driver for BF4. I'm sure nvidia will make a heroic effort on their driver, because software is an area where nvidia basically excels. I know it's early for the BF4 title and the dust hasn't settled yet, but I remember HardOCP did several reviews for BF3 since those games are played for months and years past their release. I expect the same with BF4 - i'm sure there will be a followup article for BF4.

 

[Silverforce11]

I always find it odd some people rabble on about "whats the point of doing MP tests??" clearly in a game that is going to be 99.9% of the time, played in MP by people who buy it.

Lots of sites are slack, they aren't willing to spend hours gaming on each card in MP and form an accurate impression. [H] isn't one of these sites which is why its still highly respected by gamers.

 

[BrentJ]

http://hardforum.com/showpost.php?p=1040391945&postcount=78

 

[BallaTheFeared]

So no followup with the 780 Ti, and probably nothing in the way of OC vs OC?

Disappointing, would have like to see SP numbers as well to get a baseline.

 

[Silverforce11]

There's heaps of other sites that have done SP if thats what you're after.

 

[BallaTheFeared]

Different setups, hard to gauge.

I haven't bought it, waiting for the big price drop and mantle.

Still would like to see H OC the 290x and 780/Ti, despite my problem with MP I do still enjoy their reviews.

 

[BallaTheFeared]

Dang, that bad?

http://www.youtube.com/watch?v=MknLARblrDQ

 

[Silverforce11]

Dang, that bad?

http://www.youtube.com/watch?v=MknLARblrDQ

Yup, buddy of mine got it at launch, and its got graphics glitches and sound bugs (doesn't like fancy sound cards).. others have random crashes. I am not getting it til its patched up and fixed.

 

[crisium]

The crashed are disappointing. As are other glitches.

But I still have fun.

Maybe I'm part of the problem, but as long as DICE/EA eventually fix it, I am fine with it. I'd rather play a Beta with 100% maps and weapons in October-November-December then wait months for a fully working version. Either way I'm paying nearly (GMG discounts) full price.

As long as they eventually fix it, that is. Otherwise, while I still am having fun (48 hours already), I will feel like a bought a broken product. But they should get it working eventually.

 

[Schro]

My ears are ringing. Figure I should swing by here to say hi.

I think the point is that the anomaly shouldn't exist. It raises the question about possible other anomalies that might skew the other results as well. I'm actually surprised they left the 760 score like that and didn't try and get a better representative score.

Unfortunately, by the time I realized that the 760 score was higher than the 770, the client side patch from last Friday had already dropped, so going back to rerun the test for better representative data was not going to be very possible due to the wide variety of changes made in the patch.

Quite frankly, this is a very difficult game to benchmark and to get data that is representative of the game play experience. Even when you try to do the same general tasks in a map to try to get a level performance, it is difficult to get the other 63 players to cooperate. Some matches were literally tank shootouts with explosions galore while others there was nary a tank in sight. When you multiply that by the bug in the Siege of Shanghai map that dumps 2/3s of the players out of the server with a crash when the building at C is demolished, lets just say there were many rage moments doing the data collection for the article.

As a note on the driver for NVIDIA, the (unreleased) 331.70 Beta that we were using is supposed to be the game-ready driver for BF4. The latest release in the wild right now is 331.65.

 

[Arkaign]

AMD is still better for BF4 for sure.

That said, they REALLY need to figure out an apples/apples MP testing.

I propose empty server passworded + a few dozen dummy accounts with scripted controls running around shooting buildings or whatever.

I mean I know it would entail actual work, but this test is BADLY botched by the randomness of the game.

We know that 290X > 780 (and most likely 780ti) with current drivers. Etc.

But 760 is NOT > 770. By this review, it's hugely faster. Yes, they can make notes for that one comparison. But the truth of the matter is that EVERY game of live MP is going to be different. Even on the same 64P server with the same map with 64 players. Depending on where you go, what's happening around you, and what each of the other players is doing, things are going to be vastly VASTLY different. WAY too different for you to be able to tell what the fork is up with the benching.

Want to know how I know 290X is faster at BF4? SP benches.

 

[Silverforce11]

MP is very variable, thus it requires many runs through to get a good snapshot, an average of 10 games on a 64 player map would likely yield a very accurate representation.

This is why few sites do it at all, doing 1-2 run through is highly misleading.

 

[3DVagabond]

My ears are ringing. Figure I should swing by here to say hi.



Unfortunately, by the time I realized that the 760 score was higher than the 770, the client side patch from last Friday had already dropped, so going back to rerun the test for better representative data was not going to be very possible due to the wide variety of changes made in the patch.

Quite frankly, this is a very difficult game to benchmark and to get data that is representative of the game play experience. Even when you try to do the same general tasks in a map to try to get a level performance, it is difficult to get the other 63 players to cooperate. Some matches were literally tank shootouts with explosions galore while others there was nary a tank in sight. When you multiply that by the bug in the Siege of Shanghai map that dumps 2/3s of the players out of the server with a crash when the building at C is demolished, lets just say there were many rage moments doing the data collection for the article.

As a note on the driver for NVIDIA, the (unreleased) 331.70 Beta that we were using is supposed to be the game-ready driver for BF4. The latest release in the wild right now is 331.65.

Thanks for stopping by and explaining. Kind of sounds like the whole idea was an exercise in futility. Too bad that you had to use data that some people would find questionable. I'm sure you realize that the fans of the "losing" card are going to try and find any reason to dismiss the whole yest out of hand. That's just the nature of enthusiasts, though.

Can't wait for Mantle. You ought to see if you can get on beta testing. Obviously so you can leak info to us.

 

[Arkaign]

MP is very variable, thus it requires many runs through to get a good snapshot, an average of 10 games on a 64 player map would likely yield a very accurate representation.

This is why few sites do it at all, doing 1-2 run through is highly misleading.

Even 10 runs can be wrong though.

Situation :

8 rounds on each are purely average.
2 rounds on card A are heavier than normal due to server lag / glitching / super coincidences (squad all using big vehicles, blowing up lots of enemy vehicles in view of player, etc)
2 rounds on card B are lighter than normal due to X, Y, and Z (eg; a squad of players taking a break to go take a dump, eat, whatever, being idle)

That could seriously cause card A to look slower than it is on average, and card B to appear faster.

This is why the only thing to do is bots + scripting + locked server. For the test PC, it wouldn't be able to tell the difference between a real match and the test match.

Given how Origin and EA/Dice work, that's probably unfeasible. But until it is, take any MP testing with an absolutely epic grain of salt.

Cliffs : want to know which is the fastest video card list for BF4? Look at SP. MP will be heavier, but it won't cause cards to flip around in position nonsensically. 290X is still fastest, etc, along with 770 > 760 (derp).

 

[Arkaign]

Counter-idea.

In the case of testing only two cards, have two otherwise identical test systems join the same server on the same team at the same time. Go to side of map and watch from a distance side by side as close as possible.

Not perfect, but far more representative than trying to randomly go through games. I've played probably 10,000 rounds of BF2,3,4, and no two games are the same.