AMD Catalyst 14.2 Beta V1.3 Now Available

Interesting...you are the 4th person (the other's being on the BF threads) that have said the same thing, after first clean install the game is performing spectacular, subsequent reboots either reset settings or we are changing something on our systems in between that corrupt...something. I noticed a HUGE performance decrease after the second boot, as have others. So frustrating; and I totally agree, fatigue is setting in. It really is a FUN game, but it would be so much more engaging if it ran correctly =( I am still of a mind to sell these and grab the 780, but I'm also a tinkerer, I just love new tech and tweaking the hell out of my system, so Mantle is fun, to a point.
 
ive noticed performance decreases after remaining idle for a period of time.

and my damn clocks keep sticking in 3d mode after coldboot.
 
While there's no Mantle love for 270X yet but DirectX plus 8 threads on a 16-core getting ~40fps at 1080p ultra 4xaa on demanding maps like multiplayer Siege of Shanghai. Once Mantle comes around I'm hoping it'll be close to 60fps.

Performance is consistent after reboots with the 290X with Mantle and virtually 100% stable except if I switch the video settings too many times in game while configured for Mantle it causes BF4 to crash when quitting the game. No big deal as long as it doesn't happen in-game. DirectX doesn't appear to have this issue.
 
Last edited:
Mugato said:
Interesting...you are the 4th person (the other's being on the BF threads) that have said the same thing, after first clean install the game is performing spectacular, subsequent reboots either reset settings or we are changing something on our systems in between that corrupt...something. I noticed a HUGE performance decrease after the second boot, as have others. So frustrating; and I totally agree, fatigue is setting in. It really is a FUN game, but it would be so much more engaging if it ran correctly =( I am still of a mind to sell these and grab the 780, but I'm also a tinkerer, I just love new tech and tweaking the hell out of my system, so Mantle is fun, to a point.
Click to expand...

To be realistic Mantle is rather new. Expecting Mantle to work flawlessly after 2 driver releases is not realistic.
 
Can anyone confirm pros, cons of playing around with CPU threads.
I have not had time to play around with it yet. + where are the Nvidia funboys :p
 
BF4 will crash most of the time if your GPU is OC'D too much, including memory.

Make sure to find the sweet spot for stability, Mantle is making the GPU do far more work than DX.
 
TroyX said:
BF4 will crash most of the time if your GPU is OC'D too much, including memory.

Make sure to find the sweet spot for stability, Mantle is making the GPU do far more work than DX.
Click to expand...

This is true. I spoke to TroyX earlier today and he recommended I lower my clocks, due to the issue of bf4 freezing up my system after some time of gameplay. I lowered my core clock by 10 MHz, which is down to 1060mhz on my two 290s. All has been fine so far.

His theory makes a lot of sense...much like how a cpu can be stable for daily use, but when you encode and use 100% CPU, you end up experiencing a crash.
 
Pieter3dnow said:
To be realistic Mantle is rather new. Expecting Mantle to work flawlessly after 2 driver releases is not realistic.
Click to expand...

It would have been nice if they waited to launch until they supported more products. For two months, all they have (barely) officially supported is a GPU that is owned by maybe 1% of AMD users...


*After one of the patches in January, BF4 started to bluescreen about every 40 minutes. I had to lower my CPU and Northbridge overlcocks to keep BF4 from blue-screening. 100mhz less on the CPU and about 300mhz less on the northbridge (Phenom II's love overclocked northbridges). However, My videocard overclock has no negative effect.
 
Blazemore said:
Does AB work fine with the new 14.2 drivers? I couldn't get AB to maintain a stable oc with the 14.1 and went back to 13.12.
Click to expand...

Keep in mind that it could be that the GPU is stressed more with Mantle and the O/C just might not be as stable as you thought it was.
 
Violator said:
More positive so far than 14.1. I don't appear to have the artefacts in the toolbar and missing text in webpages anymore.
Click to expand...

Same here, no crashing at all so far (only used it for a few hours. On the 13.12, 14,1 I could not use any flash/air based apps. Had tons of artifacts on the desktop and random lockups.

Haven't tested any games yet but will do tonight.
 
mesyn191 said:
Waiting can't help since Mantle is dependent on certain hardware features to work which their older non GCN GPUs don't have.
Click to expand...

Not only that but since these are beta drivers, can you imagine if they launched it to an ass load of their products at the same time? Iinstead of just having to fix drivers for a product and get it working, they would be slammed trying to fix them for a multitude of products. In turn pissing off even more people than just the select products the beta was meant for.
 
mesyn191 said:
Waiting can't help since Mantle is dependent on certain hardware features to work which their older non GCN GPUs don't have.
Click to expand...

Right. But I wasn't referring to those GPUs. The only GPU they officially support right now is the 290/290x *and Kevari. Which maybe 1% of AMD users have. A lot more people have HD7 series cards (which are GCN 2.0) and the rebaged cards like the 260/270/280.

the GCN 1.0 cards/HD6 series remains a mystery as far as Mantle and AMD seems to be avoiding talking about it for now. So it's not clear what type of benefits those cards might see.

Drep said:
Not only that but since these are beta drivers, can you imagine if they launched it to an ass load of their products at the same time? Iinstead of just having to fix drivers for a product and get it working, they would be slammed trying to fix them for a multitude of products. In turn pissing off even more people than just the select products the beta was meant for.
Click to expand...

I really didn't think I would have to spell this out but:

Ok the idea was they should have waited until they were ready. AMD clearly was not ready. Note even close. If they had communicated from the start that the roll out would be something like this, I'd be more fine with it. But the impression was that Mantle was lanching in December with BF4 and GCN 2.0 cards would see big boosts (which they do!)-------and then we find out two days before the end of December that they are pushing it to January (DUH!) and then they finally launch at the end of January with a two product beta. I know what beta means and I know that "beta" has been misrepresented by AMD and Nividia for a long time, with drivers that are often very solid. But they really should not have launched on a beta. Especially after they gave us the impressions it was going to be totally awesome.

They missed the launch by a whole month (I'm surprised I didn't hear about their stock taking a dive). Then they barely supported two GPUs, with a driver that for many, wasn't even a usable windows driver. Ok, the mantle side of the driver is beta-----but garbled text and choppy internet video playback in windows???? and the initial driver files were missing the driver I.D. string and often failed with corrupt files???. 1 month later, they fix the windows side of the driver. But they still only support one GPU. And it took DICE 1 month to fix what was said to be a simple rendering bug in Mantle (brightness/fogging issues).

I mean what a completely flat impression to give for your huge new API. I like AMD, I really do. And I have had AMD and ATI stuff for years. But this has been a really poor launch for Mantle.
 
Last edited:
chameleoneel said:
Right. But I wasn't referring to those GPUs. The only GPU they officially support right now is the 290/290x *and Kevari. Which maybe 1% of AMD users have. A lot more people have HD7 series cards (which are GCN 2.0) and the rebaged cards like the 260/270/280.

the GCN 1.0 cards/HD6 series remains a mystery as far as Mantle and AMD seems to be avoiding talking about it for now. So it's not clear what type of benefits those cards might see.
Click to expand...

Um, no. The HD6xxx series is not GCN 1.0. GCN 1.0 was introduced with the 7xxx cards.
 
What do you mean it is supporting one GPU. It works on 7970 as well.
Lots of pessimistic views here. I thought we were discussing performance gains/tweaks. The fun to tinker with this new stuff. Yes its only BF4 for now, but its new tech its exciting to say the least. Free performance gains. The potential of it in the long run. It looks like some of us here are relapsing and are interested in critique and judgment of how poorly it was released.
 
are you really surprised all the haters have come out of the woodwork?

its pretty much rote at this point that any technology from either side that potentially makes your latest purchase shine a little duller brings out all the trolls.
 
No, I'm not surprised.;). Trolls will be trolls. It is fun to read how blinded by the hatred some of are.
Someone ban me lol....
 
Imhotep said:
What do you mean it is supporting one GPU. It works on 7970 as well.
Lots of pessimistic views here. I thought we were discussing performance gains/tweaks. The fun to tinker with this new stuff. Yes its only BF4 for now, but its new tech its exciting to say the least. Free performance gains. The potential of it in the long run. It looks like some of us here are relapsing and are interested in critique and judgment of how poorly it was released.
Click to expand...

They only officially support 290/290x and Kevari. For many people, the HD7 series in Mantle uses so much extra VRAM compared to DX11 for the same settings, that the framerate boost is completely offset by constant stuttering/hitching due to exceeding VRAM limits.

Ozymandias said:
Um, no. The HD6xxx series is not GCN 1.0. GCN 1.0 was introduced with the 7xxx cards.
Click to expand...

my bad


--------------

Also, I'm not trolling. I'm expressing genuine concern for the way AMD is handling/presenting Mantle in reality. I would love for it to be a raging success. Which is why I have spent a lot of personal time collecting performance data to send to them.
 
Last edited:
Welp, unlike 14.1, these actually work with 290x in CF. I'd say that performance with mantle is roughly on par with directx though. I'm running an i7 920 @ 4ghz, 12 gigs of ram, 5760x1200.

With mantle, I did get an "out of memory" issue for the first time last night though.
 
mesyn191 said:
79xx are GCN too. 14.2's are working fine for me. But then so did the 14.1's.
Click to expand...

Official AMD sources have said that the 14.1 and 14.2 drivers contain no specific optimizations for the HD7/HD79 series. So far they have only focused on Kevari and the 290/290x.

Have you compared your VRAM usage between DX11 and Mantle for the same settings? I would guess it is much higher, even if it isn't overflowing your VRAM. If it isn't much higher, well that would be interesting to know.

I'm guessing that HD79XX users reporting only good results are not hitting a VRAM wall, since all of the HD79 cards have 3gigs or more of VRAM.

In this post here, I compare VRAM usage for 14.2
http://hardforum.com/showpost.php?p=1040657036&postcount=64

and in this post here, I compare VRAM usage for 14.1
http://hardforum.com/showpost.php?p=1040594668&postcount=29

The general takeaway is:

1. both 14.1 and 14.2 show notable framerate boosts in mantle, for my everyday play settings. Even though my 7870 is apparently not yet a target for AMD.

2. The overall framerate boost is completely ruined by the fact that that there is stuttering and hitching every 3-5 seconds.

3. 14.2 reduces the severity of the stutter, but it is still a significantly negative impact on the gameplay experience. Highlighting the fact that overall frames per second readings aren't everything. A lot can happen inside of a second.

4. VRAM usage for 7850 and 7870 has been shown by users in this forum to be significantly higher in Mantle VS. DX11, with the same settings. Usually saturating available VRAM.

5. This is likely a key factor in the stuttering.
--------------------------------------------------------------------------------------

I don't recall any users with 290/x or HD79xx comparing VRAM usage. So it is unclear if this is a general bug that just isn't severe enough to affect cards with 3+ gb of ram, or if the bug is isolated to HD7 cards.
 
Last edited:
Arcygenical said:
Still getting the odd black screen in fucking windows.

This is very frustrating. Sometimes my DVI connection will just shit the bed. A hard reset often brings it back, but sometimes I need to boot my PC without the monitor cable attached, turn it off again, and re-boot with the cable attached.
Click to expand...

I found that usually a Ctrl+Alt+Del fixes it, but yeah, it's really annoying. I rolled back to 13.12 until they fix it (hopefully next beta in March).
 
chameleoneel said:
Official AMD sources have said that the 14.1 and 14.2 drivers contain no specific optimizations for the HD7/HD79 series.
Click to expand...
No specific optimizations are necessary for GCN 1.0 GPU's to benefit. Why would you think otherwise? All GCN revisions are broadly similar so not much if the way of specifics should be needed.

chameleoneel said:
Have you compared your VRAM usage between DX11 and Mantle for the same settings? I would guess it is much higher, even if it isn't overflowing your VRAM.
Click to expand...
Haven't bothered to check because everything is working fine which is all I care about. The few weird issues that occurred with the 14.1's during desktop usage are gone now. Thief gameplay seems a little smoother but that is subjective. All the other games I bother to play (DoD:S, CS:GO, Insurgency, PinballFX2, Pinball Arcade, Trine 2) work fine too.
 
I don't recall any users with 290/x or HD79xx comparing VRAM usage. So it is unclear if this is a general bug that just isn't severe enough to affect cards with 3+ gb of ram, or if the bug is isolated to HD7 cards.

No vram issues on 7970. All settings maxed out at 1920x1080.:cool:
 
Imhotep said:
I don't recall any users with 290/x or HD79xx comparing VRAM usage. So it is unclear if this is a general bug that just isn't severe enough to affect cards with 3+ gb of ram, or if the bug is isolated to HD7 cards.:
Click to expand...


No vram issues on 7970. All settings maxed out at 1920x1080.
 
Imhotep said:
No vram issues on 7970. All settings maxed out at 1920x1080.
Click to expand...

The point is that some of use would like to see data. Use something like GPU-Z. As I mentioned, the VRAM issue could be there, you just may not be feeling it with 3GB of VRAM or the added power of crossfire (not you, but I've seen some HD7/HD79xx crossfire users saying their gameplay is very smooth).

mesyn191 said:
No specific optimizations are necessary for GCN 1.0 GPU's to benefit. Why would you think otherwise? All GCN revisions are broadly similar so not much if the way of specifics should be needed.
Click to expand...

I would think so because

A. many HD7 users, as well as a couple of websites, have reported worse performance on HD7 cards and their derivatives (mobile HD8 series and the R270/x) with mantle VS. DX11. There's also been a few of us in the harddforums that have reported worse Mantle performance with 14.2 VS. 14.1

B. many users and websites have reported constant stuttering/hitching/and pauses on HD7 cards, even though overall FPS is often higher. Including HD79xx users on Guru3d forums (still no VRAM data, though)

C. some users of 7850 and 7870 have investigated further and found huge discrepancies in the amount of VRAM used for the same settings

D. the release notes for 14.1 say "Mantle performance for the AMD Radeon™ HD 7000/HD 8000 Series GPUs and AMD Radeon™ R9 280X and R9 270X GPUs will be optimized for BattleField 4™ in future AMD Catalyst™ releases. These products will see limited gains in BattleField 4™ and AMD is currently investigating optimizations for them."

Since the VRAM and/or stuttering issues are still present with 14.2 and FPS appear to be overall lower in 14.2 than 14.1, I'd wager this is still the case. Even though they removed that note from the 14.2 release notes.

there's also this bit from twitter:
https://twitter.com/TheMattB81/status/438672517927305216

although it's unclear if that guy is an AMD insider or just an passionate fan.
 
Last edited:
chameleoneel said:
There's also been a few of us in the harddforums that have reported worse Mantle performance with 14.2 VS. 14.1...B. many users and websites have reported constant stuttering/hitching/and pauses on HD7 cards....C. some users of 7850 and 7870 have investigated further and found huge discrepancies in the amount of VRAM used for the same settings
Click to expand...
That can happen for all sorts of reasons like bugs in either the game or the drivers or a combination of both.

chameleoneel said:
D. the release notes for 14.1 say "Mantle performance for the AMD Radeon™ HD 7000/HD 8000 Series GPUs and AMD Radeon™ R9 280X and R9 270X GPUs will be optimized for BattleField 4™ in future AMD Catalyst™ releases. These products will see limited gains in BattleField 4™ and AMD is currently investigating optimizations for them."
Click to expand...
The statement is regarding specific optimizations, not general purpose ones that will be effective for all generations of GCN.
 
Does anyone remember what CPU, Mobo+Chipset combinations the mentioned users of HD7xxx are running. My Set up runs fine , but I think it is safe to say that Mentle leaks memory left and right.
3 Gigs takes a little longer to fill than 7850 1 gig. What I would like to know why is it that my xfire 7970 runs fine. Was I lucky not to experiance that and by the time the Vram would be overfilled the round and map has changed. Is it my hardware and how it is configured. It would be great if we had all the set ups and configs available to compare. :)
 
Getting hardware configs will be imposible. Is there anyone in this thread running a single 290x or xfire.
Those with HD 7xxx series as well. I would like to know what is your Row Refresh Cycle Time (tRFC) set to. CPU-Z under memory settings will show that. Just those guys experiancing stutters and overfilled Vram issues.
 
mesyn191 said:
The statement is regarding specific optimizations, not general purpose ones that will be effective for all generations of GCN.
Click to expand...

I'm not sure if you like to have the last word or if you are just trolling me at this point. If you want to focus on a technicality in the wording, whatever. I think it is pretty clear by this point what I mean when I say that AMD is not supporting anything right now but Kevari and 290/290x.

Imhotep said:
Does anyone remember what CPU, Mobo+Chipset combinations the mentioned users of HD7xxx are running. My Set up runs fine , but I think it is safe to say that Mentle leaks memory left and right.
3 Gigs takes a little longer to fill than 7850 1 gig. What I would like to know why is it that my xfire 7970 runs fine. Was I lucky not to experiance that and by the time the Vram would be overfilled the round and map has changed. Is it my hardware and how it is configured. It would be great if we had all the set ups and configs available to compare. :)
Click to expand...

It's not a VRAM leak in the sense that it just keeps using more and more VRAM forever.

It's just straight up more VRAM usage for the same settings. It doesn't keep filling up. It's a static amount of ram that can be lowered with lower settings. If I drop to "medium" textures and "high" mesh quality, I can predictably drop the VRAM usage and get rid of stuttering. The problem is that the VRAM usage is so much higher for the same settings, I can no longer use my everyday play settings in Mantle. Which otherwise perform well in DX11 and only use about 1200mb of VRAM:

ScreenshotWin32-0002.png


Here's 14.2 DX11 Vram usage on 64player Siege of Shanghai, with my settings pictured above, taken some time after I ran the 6 minute benchmark

https://www.dropbox.com/s/knipayekayyldu3/14_2DX11SoS.jpg

Here's 14.2 Mantle VRam usage on 64player Siege of Shanghai, with my settings pictured above, taken some time after I ran the 6 minute benchmark

https://www.dropbox.com/s/5bs3h0hjkewj0pm/14_2MantleSoS.jpg
Click to expand...

To mitigate the stuttering, I had to drop my Textures to medium and my mesh quality to high, which is just unacceptable.

I actually took a screenshot of the VRAM usage for those setttings, but forgot to save it. it was about 1600mb usage. My 7870 has 2GB of VRAM.
Click to expand...

*full post with all of my 14.2 data so far:
http://hardforum.com/showpost.php?p=1040657036&postcount=64

Imhotep said:
Does anyone remember what CPU, Mobo+Chipset combinations the mentioned users of HD7xxx are running. My Set up runs fine , but I think it is safe to say that Mantle leaks memory left and right.
3 Gigs takes a little longer to fill than 7850 1 gig. What I would like to know why is it that my xfire 7970 runs fine. Was I lucky not to experiance that and by the time the Vram would be overfilled the round and map has changed. Is it my hardware and how it is configured. It would be great if we had all the set ups and configs available to compare. :)
Click to expand...

Please run around in the test server or a game for a few minutes, then take a screenshot showing yoru VRAM usage with GPU-Z. Do it for both Mantle and DX11, with the same settings. I would love to see it. Single card mode would be great as well!

I'm guessing that 7950/7970 also experience higher VRAM usage. But depending on your settings, you still may not be filling up 3GB of VRAM. Additionally, the extra horsepower of crossfire probably brute forces through the higher VRAM usage.


I have a Phenom II X6 running on an Asus M4A89GTD PRO/USB3 which is the 890GX chipset. HIS 7870 ghz edition. 8gigs of Samsung 1600mhz low profile ram, currently running at 1367mhz, due to having to relax my multiplier for CPU and northbridge overclocks in BF4.

Imhotep said:
Those with HD 7xxx series as well. I would like to know what is your Row Refresh Cycle Time (tRFC) set to. CPU-Z under memory settings will show that. Just those guys experiancing stutters and overfilled Vram issues.
Click to expand...

would you mind explaining what relevance the tRFC setting has with VRAM usage and constant stuttering/hitching with game performance?

My tRFC is currently set at 30. It used to be 33, but I was also running my memory at higher mhz and looser timings than I am right now. i had to relax my CPU/Northbridge/Memory overclock after one of the BF4 patches in January. As It caused me to bluescreen about every 40 minutes. Even though every other game I have played with those settings for the past year, has been stable.
 
Last edited:
Im not trolling you and my word does not need to be the last one. Just like you im interested in why it works for some and does not for others :)
Mentle was meant to give more work to the GPU's, so it allocates more memory and if it does not leak. Why do some of us experiance stuttering.
Dont take this too seriously. We are just thowing ideas around. Im with you on this.
 
Last edited:
Imhotep that response wasn't directed at you. If you double check my post, you'll see I quoted someone else and was responding to them.

*I updated with hardware specs and tRFC
 
Last edited:
My Apologies. Before mantle lots of people were complainig about microstutters in DX11 on xfire set ups. Mine were not as bad as others. What I found is Row Refresh Cycle Time (tRFC) if run to agressively will make the game stutter. Your finding proves that it will also bluescreen.
Amd's drivers had never anything to do with stuttering. The moment I relaxed Row Refresh Cycle Time (tRFC) on my setup BF3 and 4 are silkly smooth in DX11. My guess is that Mantle is a simular story :)
 
Imhotep said:
My Apologies. Before mantle lots of people were complainig about microstutters in DX11 on xfire set ups. Mine were not as bad as others. What I found is Row Refresh Cycle Time (tRFC) if run to agressively will make the game stutter. Your finding proves that it will also bluescreen.
Amd's drivers had never anything to do with stuttering. The moment I relaxed Row Refresh Cycle Time (tRFC) on my setup BF3 and 4 are silkly smooth in DX11. My guess is that Mantle is a simular story :)
Click to expand...

So what's your tRFC? and would you mind using GPU-Z to screenshot some VRAM usage comparisons, similar to what I posted above?
 
No, I don't mind, but it will take a bit of time since my 4 year old is assisting me :p
Im using a core I7 980x at 4.333 ghz with 12 gigs of ram. Memory at 1600 9.9.9 Row Refresh Cycle Time (tRFC) at 128.Asus Rampage III Extreme.
Im aware that AMD chips run tighter Row Refresh Cycle Time (tRFC) :)
 
Last edited:
I just wanted to throw in my 2 cents.

R9 290 (1075/1400) + 3770S (3.1 ghz)

BF4 - Never dips below 50fps, typically around 80.

Resolution: 6040x1080 (eyefinity)
Texture Quality: Ultra
Texture Quality: Ultra
Lighting Quality: Medium
Effects Quality: Medium
Post Process: Low
Mesh Quality: High
Terrain Quality: High
Terrain Decoration: High
AA: Off
Post AA: Medium
Ambient: SSAO

The 14.1 Betas were honestly a pain in the ass and would crash while loading up the game 50%. These are completely stable for me and I see no issues at all. Additionally, there are no longer the FPS hiccups that would previously occur.

For performance comparison using DX:

Resolution: 6040x1080 (eyefinity)
Texture Quality: Ultra
Texture Quality: Ultra
Lighting Quality: Medium
Effects Quality: Medium
Post Process: Low
Mesh Quality: Medium
Terrain Quality: Medium
Terrain Decoration: Medium
AA: Off
Post AA: Medium
Ambient: SSAO

My FPS were significantly lower. Min FPS would frequently hit 35 fps but would hover around 50-60.

For me, Mantle has made a HUGE performance difference. Not only is it smoother but I can even bump up a few more settings. It was very buggy at first but after one driver revision, it's been completely stable and an extremely smooth experience. Also worth noting, like others, 50 fps in Mantle vs 50 fps in DX seems like a different experience. It's simply smoother in Mantle.
 
Rizen said:
Yes, I'm using AB with 14.2.
Click to expand...

Odd, switch over to 14.2 and AB will not keep a constant core oc. In heaven its going up and down on the core. With 13.12 I can keep a constant flat line 1200 core.
 
You must log in or register to reply here.
Back
Top