Mantle Feedback/Bugs

Rusty0611 · 3 Feb 2014 at 23:04

pgi947 said:
The point I was making was about cpu usage, if you'd care to read back your own comments, at the bottom of my post I said I would get around to a stock run, my whole point was based around the benches you posted showing an absolute terrible average on both dx and mantle with a 290x on a stock 4770k.

Anyhow, can do without trying to put the effort in to get something blindly dismissed.

**** this I'm out, totally, these forums have become a hell hole over the last few months.

Later fellas.

It's all good mate - I'm sure myself and others will take the value in it by understanding the statistical representation of ya results

LtMatt · 3 Feb 2014 at 23:05

pgi947 said:
Their testing must be a little skewed or something in that case, Shanghai, 64 player (at least 95% populated) 1920x1080 full ultra with 4770K at 4.6ghz Single GTX780 @ 1006/6000 around 15 mins testing:

Frames, Time (ms), Min, Max, Avg
77770, 904468, 58, 113, 85.984

If I get chance I will re run with CPU at stock and see what differences I get.

pgi947 said:
The point I was making was about cpu usage, if you'd care to read back your own comments, at the bottom of my post I said I would get around to a stock run, my whole point was based around the benches you posted showing an absolute terrible average on both dx and mantle with a 290x on a stock 4770k.

Anyhow, can do without trying to put the effort in to get something blindly dismissed.

**** this I'm out, totally, these forums have become a hell hole over the last few months.

Later fellas.

Didn't mean to blindly dismiss, perhaps i could have phrased it better, but i feel you're overreacting somewhat. I was replying to you dismissing their results and was just saying why it might not be an accurate like for like comparison. Their benchmark involves a 30 second run looking across the whole city. Your benchmark was 15 minutes long, hence my point about not an accurate comparison to what they benched. Anyway sorry if this offended you somewhat. You can see their benchmarking run by clicking this link and watching the video at the bottom.

http://www.pcgameshardware.de/AMD-R...ecials/AMD-Mantle-Test-Battlefield-4-1107754/

Rusty0611 said:
Which is the whole point that if it is representative and long enough then it doesn't matter that it isn't identical.

See above. That's why its largely pointless. Not meaning to knock pgi at all, but he dismissed their benchmark results so i explained why his comparison was largely useless. Nothing wrong with him benchmarking and i trust his results. People just need to look what i actually said.

tommybhoy · 3 Feb 2014 at 23:06

All the oCuk BF3 runs for the BM thread were done on Operation Swordbreaker single player in BF3 to the point you all hated doing the same run through.

No point changing the goalposts bhoys.

setter · 3 Feb 2014 at 23:08

LtMatt said:
Exactly. It depends how the game was played and scene was benched. Going round in a chopper or sniping on a roof = highest fps. Spending more time in a tank, lowest fps. Running round on foot, somewhere in the middle but generally higher fps. That's why its not reliable comparing multi player benches. Its only representative if both players do the same thing. Also depends on how long each bench is run for.

Tbf, myself, pgi and thefogo have been running a few benches this evening in the bf4 thread on here. We all have pretty similair systems and benched on different servers at different times. The fps figures are pretty similair for all three of us, particularly minimum and averages.

Gregster · 3 Feb 2014 at 23:08

pgi947 said:
The point I was making was about cpu usage, if you'd care to read back your own comments, at the bottom of my post I said I would get around to a stock run, my whole point was based around the benches you posted showing an absolute terrible average on both dx and mantle with a 290x on a stock 4770k.

Anyhow, can do without trying to put the effort in to get something blindly dismissed.

**** this I'm out, totally, these forums have become a hell hole over the last few months.

Later fellas.

LtMatt · 3 Feb 2014 at 23:10

setter said:
Tbf, myself, pgi and the fogo have been running a few benches this evening in the bf4 thread on here. We all have pretty similair systems and benched on different servers at different times. The fps figures are pretty similair for all three of us, particularly minimum and averages.

Yep not doubting the results. People are too quick to overlook the reasons for why comparing the two results were largely pointless though. Ah well another day on ocuk.

GoogalyMoogaly · 3 Feb 2014 at 23:10

I think this is the problem with benchmarks, the ones that can be repeated are the least useful.
Unless you're only interested in playing the level Matt suggested then that benchmark might not be that representative of a normal user's experience. But might be repeatable, well more closely than a multiplayer run.
I think it's fair to say most people will play multiplayer the most. Maybe this is where I differ, but when I play multiplayer I don't do the exact same thing on a map every time I play it. So my experience one day might not be the same as my experience another day. So in that way doing multiplayer benchmarks may actually be more representative of the gameplay experience. Doesn't make for good graphs though.

Now I realise we're all looking to draw some pretty graphs so we can have some proof that a 290X on Mantle ridicules a 780, but maybe we should wait until we have proper benchmarks we can run. Like 3D Mark or Heaven type things.
Wait until the Star Swarm demo has a benchmark option (I thought they said they were working on it?) or see if Thief does.

Until then these benchmarks don't really prove much, unless everyone records their benchmark run there's no telling how they played even the single player level. Maybe they ran it enough that they could spend a lot of the time looking at the floor and skew the results. Or perhaps different playstyles would cause this to happen naturally.

We waited well over a month for Mantle past it's supposed release date, can we not just wait a bit longer for some actual benchmarks (standalone or built into a game)?

Rusty0611 · 3 Feb 2014 at 23:21

LtMatt said:
See above. That's why its largely pointless. Not meaning to knock pgi at all, but he dismissed their benchmark results so i explained why his comparison was largely useless. Nothing wrong with him benchmarking and i trust his results. People just need to look what i actually said.

It isn't though. As I have said, if the run is long enough and is representative of BF4 gameplay in general then it is comparable.

andybird123 · 3 Feb 2014 at 23:25

Rusty0611 said:
It isn't though. As I have said, if the run is long enough and is representative of BF4 gameplay in general then it is comparable.

Your not conforming to his agenda rusty, so that makes your point irrelevant

DamnedLife · 3 Feb 2014 at 23:28

Do you guys know about this https://www.amd.apply2jobs.com/Prof...ion=mExternal.showJob&RID=28367&CurrentPage=1 even if you do, nice to clear the air of some tension

Rusty0611 · 3 Feb 2014 at 23:30

andybird123 said:
Your not conforming to his agenda rusty, so that makes your point irrelevant

It's 'you're' but yes I think that you're right. Lub you Matthew

.

GoogalyMoogaly · 3 Feb 2014 at 23:45

Maybe we could have a separate Mantle BF4 Multiplayer thread where benchmarks are done in multiplayer?
I mean we already have BF4 benchmarks in the BF4 benchmarking thread, the Mantle benchmarking thread and the old Mantle thread, what's one more? especially if it shows something else and Matt won't allow it to be done in his threads.

EDIT:
I should probably point out that I'm not going to do either as I think they're both pointless from a benchmarking point of view that's looking to get stats.
I think there's a reason 8Pack goes for 3DMark world records and the like but we've not seen him win any world records for a best score in BF3 when playing on level 6 on easy...

DavieJG · 4 Feb 2014 at 01:17

Pretty impressed with mantle thus far.

i5 750 3.8GHz
7970 1100/1500
8GB RAM
Win 7
2560x1440

Seeing +100% on average fps in star swarm 21->45 and +200% on 1% time 4->12.
Seeing around +33% on BF4 multiplayer 75->100 (custom settings)

Only gripe is that the CPU hitching/spiking for a couple of frames every few minutes makes it a no go on BF4 for me at the moment though. There's no evidence of this in star swarm though so fingers crossed it will be ironed out in the next patch or driver.

Jasonst888 · 4 Feb 2014 at 06:57

Well my mantle somehow managed to break itself yesterday. It froze within 30 seconds of starting operation locker every time. tried multiple fixes and restarts no joy. Changed back to DX and my frames dropped to an average of 30fps (norm is around 80) In the end i changed back to the old drivers and fps is back to normal. Not sure what happened as it was fine the day before (aside from the odd cpu spike here and there)

DamnedLife · 4 Feb 2014 at 07:20

Jasonst888 said:
Well my mantle somehow managed to break itself yesterday. It froze within 30 seconds of starting operation locker every time. tried multiple fixes and restarts no joy. Changed back to DX and my frames dropped to an average of 30fps (norm is around 80) In the end i changed back to the old drivers and fps is back to normal. Not sure what happened as it was fine the day before (aside from the odd cpu spike here and there)

Did it white screen? For me it white screens for every level within 10 seconds or so.

BuzzLloydyear · 4 Feb 2014 at 07:23

DamnedLife said:
Did it white screen? For me it white screens for every level within 10 seconds or so.

White screen is happening to me now too.

Jasonst888 · 4 Feb 2014 at 07:48

DamnedLife said:
Did it white screen? For me it white screens for every level within 10 seconds or so.

No white screen it just froze, and there was a line of static/ artifacting across both of my monitors for a second. Very strange never seen that before.

DamnedLife · 4 Feb 2014 at 07:53

It froze for me too then when Ctrl+Alt+Del it white screens. When it froze, sound goes off a few seconds later too, only way to escape is task manager and then ending task so that's why I said white screen. It is the same problem basically.

LtMatt · 4 Feb 2014 at 08:09

Rusty0611 said:
It isn't though. As I have said, if the run is long enough and is representative of BF4 gameplay in general then it is comparable.

andybird123 said:
Your not conforming to his agenda rusty, so that makes your point irrelevant

Rusty0611 said:
It's 'you're' but yes I think that you're right. Lub you Matthew .

Its not an agenda, i was not the one dismissing the review as skewed. I was merely pointing out why its unreliable to compare the benchmark they ran to the one he ran. Its really simple stuff and im surprised people have trouble grasping that.

SiDeards73 · 4 Feb 2014 at 08:15

I had dubious FPS in Mantle on my 4770k and 290 both at stock, would usually sit around 55-60fps. I had selected ULTRA settings but i think i may have manually pushed something further up, possibly Post AA or something as i know MSAA was x4.

Anyhow i downloaded the Raptr thing and let it set my game to "Quality" setting, re-run BF4 and now im getting around 72fps at stock, i pushed my 290 to 1100/1250 and i get over 100fps, mostly sits around 110fps, also the cpu was pushed to 4.5ghz but that supposedly shouldnt affect it as much?

Anyhow dropped the cpu back to stock and left the card at 1050/1250 and i get a steady 80 or so fps. Im pressed as its a lot smoother than it was previously, so much so that i can actually shoot people now with a sniper rifle