AMD Catalyst 14.12 Omega WHQL (14.501.1003.0 November 20)

Anyone else notice how the latest driver introduces a red tint into the gamma ramp for RGB values in the range ~190-255? Older driver (13-12) is fine.

 

I play bf4 r9 280x on high preset,1920x1080 no AA,20-25 min maximum,on medium 30 min,on low 40-45 min.Your mantle is eating all 3gb of vid.ram at the end the game is stopped with driver error 4101.:bang:

 

I run bf4 on high with mantle on r9 280x crossfire 2560x1440 @ 96HZ with some settings on ultra and I get no frame drops or crashes.

I treid ultra but the game works and then will lag so much that my FPS will go to 20fps with stutter every 2 seconds to 0 FPS and back to 20fps so yeah mantle still doesn't work as it should.

So in conclusion Mantle still not fixed in frostbite engine. If there was no memory leak it wouldn't try to use more VRAM than you really have when running graphics to high. I tested BF4 and Dragon Age same issue

Also not trying to be funny but why do you run 12GB ram? i think you should change it to 16GB or something that's dual channel mode.

 

I don't seem to get the memory leak in mantle on my 280x on BF4, i do however have issues with the graphics on certain levels in the SP campaign. On the 3rd level (i think its the 3rd - the assault with the tanks during the storm etc) the rain appears as long black streaks and doesn't look right, on DX11 it looks perfect.

Its very weird, i know for a fact its not an overheating issue (this has been tested a lot) and its very distracting, anyone else have issues with certain graphical effects in BF4 using mantle?. Like i said, on DX11 it looks pretty dam stunning, yet on mantle is just looks weird due to having black rain.

 

Just tried HL2 FakeFactory MOD v12 @ 2560x1440, seems to run at 100Hz, as was locked at 99.9fps, well, till i turned the flashlight on, as it would crash to around 50fps with shadows , looked bloody awesome though

 

Mantle+Crossfire works fine for me.

 

Since there are folks claiming crazy things such as mantle not supporting GCN 1.0, or that the Mantle implementation in BF4 is broken for everyone with memory leaks...

I feel compelled to mention that I can play BF4, using mantle, using my GN 1.0 card, with zero memory leaks. Everything on ultra, except just 2xAA and no motion blur. Sure, my vram can sometimes peg out at 29xxMB but it does not seem to introduce any impairment to gameplay. Usually vram usage is around 2800MB.

Strange artifacts though, that have remained in bf4 since launch? Definitely.

 

Your RAM seems pretty low and DDR2, you don't get any fps drops?

Cause i know alot of ppl have issues with 4gb ram and BF4.

 

Indeed it is, some of my hardware is even 7 years old. I do get the odd CPU spike here and there but the game plays very fluid, often around 70fps. 64 player servers are what I enjoy, but they can also be pretty hard on the CPU.
Call it swagger (/s), call it luck, call it poor guy that can't buy an i7 - I game with the best of them at almost always at max settings @ 1200p.
Makes me wonder what some of these guys with uber rigs are doing incorrectly.

 

I remember playing BF4 on Q9550 and 4GB RAM, it wasn't pretty on my rig.

Can't remember wich video card i had back then, i think it was a GTX560 Ti.
It wasn't the videocards fault, cause a friend of mine plays fluently on his with same CPU as me.

 

Well, I would literally cry if I had to play on a 560ti.. pretty sure that would be a major issue. lol
Was the Q oc'd? They don't really compete until after 3.6GHz or so these days. I'm actually looking to throw another Xeon in this board and try for 4.2Ghz.
(PS, this is the nicest discussion I've had on these boards in a looong time).

 

Yeah it was OC'd to 3,5ghz.

I played alot on my friends PC, his 560 TI is still very much capable.
Ran it on high settings, 40-50fps@1680x1050 no AA.

 

I wish there was a repeatable benchmark that I could use. I'm interested enough now though that I'm going to go test some shiz out. Got a map known for killing pc's? Otherwise I'll just jump onto hainan or zavod to mess about.

EDIT: Just played a couple rounds in a 64 player flood zone game. Ran it on ultra settings, 55-72fps@1920x1200 2xAA. Damn cpu spikes

 

You aren't getting stable 70FPS + on 1200p with Ultra + 2XMSAA on 64P maps other than Metro. I have the same MSI 7950 at 1100/1300 and I can barely pull 50 FPS on large outdoor maps when things get messy. Even Lockers in snow drops to <60s.

I do play at 2560x1080 @ 75hz (give or take 15% more pixels vs 1920x1200), but everything is on High with no AO and no MSAA. GPU load typically 80-90% when pegged maxed at 75FPS in Metro, will go to 95+ when FPS falls, power limit +20%, <65C.

Same graphics card, same settings on an i7 950 @ 3.8 and it would often have dips into 40 FPS. A8-5600K with same card and monitor is unplayable with constant dips to 20s FPS.

Therefore, I really doubt a CPU as old as the C2Q @ 3.6 can reasonably run 64P maps. I'm not sure about RAM since all the systems have at least 8GB. This is also on both W7 and 8.1 with either DX11/Mantle.

I do get the occasional stutter but I think that is more related to the game itself than either API, but Mantle seems to get hit with it much more than DX11 in my experience. I record with OBS on DX11 so I can't say for sure if the occasional lag is coming from that versus Mantle's supposed memory leak.

edit: keep typing dx12.

 

100% correct that my fps are not stable at 70. I just tested and found to be in the 55+ range when **** is hitting the fan, and anywhere up to 90 depending on what else is going on, but most often within 55-72fps. (133fps staring at the sky doesn't count). I'm sure the fps counter has hit the 30's or 40's before but it would have been an extreme circumstance that I almost never encounter.

One thing I should mention is that my experiences were not ideal on win7. Migrating to 8.1 showed significant improvement in my case.

It's tough to compare our situations since you run 15% more screen space, and different options but it does sound strange that an i7 of any generation could nose dive to 40fps, but that 2560x1080 res likely doesn't make it easy.

Any stutters I get are due to disk swapping I suspect. There is some swapping/hard faulting going on but it's not detrimental, yet. 4GB really is at it's limits though.

I guess it'd be disturbing to also learn that I seem to have FC4 running with more success than some, too?

I don't actually find my results to be all that impressive, just quite good and even competitive for aging gear.

EDIT: I've also found editing the vram timing tables to net a few more fps.

 

When checking the perfoverlay, I think most of the time it is the 7950 being maxed out long before this new 5820k did, but on the 950 and 5800, the CPUs were almost always pegged at max (the 5800 was 100% maxed always). That was why I said I find your situation a little unlikely. My friend and roommates both have 2500Ks (to rule out HT), one oc'ed and one at stock, both machines get relatively high CPU usage on 64P + large map (DX11 though, both Nvidia cards).

I would love to see a perfoverlay shot or a FCAT log though (and then find out what is wrong with AMD APU I have). I do suspect however, that the level of stuttering and FPS dips perceived is largely dependent on each person, so for example, even if there are equal amounts of spikes and FPS, I might think I'm getting worse performance (though the FPS counts I say are actually from the in game counter when things get chaotic).

I've played this game on both HDD and SSD, and I don't think its down to cache issues. The CPU spiking every now and then has only happened after the Fall? or whatever patch they did after summer. My experience was almost flawless on the old i7-950 machine other than on some maps, though I couldn't record gameplay without suffering a 10-20FPS penalty (CPU encoding). I've recently switched to using OBS with VCE, but I think with 6c and a much better OC versus the 950, even CPU encoding should be fine.

What settings do you have atm?

I do agree its very respectable performance if it can maintain 50+ on most maps and 64P.

 

I know of one wich made my FPS go down more then the others, i think it was Hainan Resort.

And i also think Gulf of Oman is pretty hefty.

 

Just tested on a ~50/64P TDM, 5ms ping, Flood Zone for roughly 5 minutes.
90 FOV, 2560x1080, everything High BUT Mesh Ultra, no AA/AO, DX11. 14.12

With 2ms smoothing



Some minor drops to 50s, mostly mid 60s. This is only TDM too. I'll try Mantle tomorrow.

 

One word: NEAT!
I want to try and put something like that together.

As for my timings, I will literally have to scour through the backwards facing hex again to tell you. But essentially, I borrowed from a few of the mining guys and changed a couple bits to better suit my needs (gaming vs. crypto hashing). I do know for sure that it is awfully close to the stock timings that already exist in the bios, it's just that I've tuned them to work for 1500MHz instead of 300 or whatever the fastest strap is. I did require an extra .05v on vram to make stable.

EDIT: mmm! http://www.overclock.net/t/1530583/fta-frame-time-analyzer-v1-0-1-supports-bf4-civ-be-da-i

 

Yep just use the tool, its super fast! Ah, I probably won't try to bother with the timings then, I remember reading about that a long time ago and it was a massive pain to test stability for.