The AMD Mantle Thread

[Arkaign]

I don't feel like reading nearly 300 pages of posts.

Did this happen or no?

http://attackofthefanboy.com/news/radeon-r9-290x-mantle-api-ridicule-titan-gpu-bf4/

 

[footballrunner800]

I don't feel like reading nearly 300 pages of posts.

Did this happen or no?

http://attackofthefanboy.com/news/radeon-r9-290x-mantle-api-ridicule-titan-gpu-bf4/

no

 

[bystander36]

I don't feel like reading nearly 300 pages of posts.

Did this happen or no?

http://attackofthefanboy.com/news/radeon-r9-290x-mantle-api-ridicule-titan-gpu-bf4/

That depends on the CPU used in the comparison, and whether or not MP is used as well.

With a top of the line CPU, OC'ed, Mantle doesn't hardly do anything for you, unless you are CF'ed, but with a slow CPU, Mantle does a good job at removing CPU bottlenecks.

 

[Arkaign]

That depends on the CPU used in the comparison, and whether or not MP is used as well.

With a top of the line CPU, OC'ed, Mantle doesn't hardly do anything for you, unless you are CF'ed, but with a slow CPU, Mantle does a good job at removing CPU bottlenecks.

Sounds like a good cliffs for me, thanks!

I have a gaming PC, 4.6Ghz 3770K, 16GB 2400, 780ti. Then I have a light gaming HTPC, FX6350 (stock), 8GB 1866, 7870.

So in short, I shouldn't care about the 3770/780ti system, but consider final mantle for the HTPC?

 

[Deders]

Sounds like a good cliffs for me, thanks!

I have a gaming PC, 4.6Ghz 3770K, 16GB 2400, 780ti. Then I have a light gaming HTPC, FX6350 (stock), 8GB 1866, 7870.

So in short, I shouldn't care about the 3770/780ti system, but consider final mantle for the HTPC?

So far it looks that way but it's early days yet.

 

[itsmydamnation]

Sounds like a good cliffs for me, thanks!

I have a gaming PC, 4.6Ghz 3770K, 16GB 2400, 780ti. Then I have a light gaming HTPC, FX6350 (stock), 8GB 1866, 7870.

So in short, I shouldn't care about the 3770/780ti system, but consider final mantle for the HTPC?

on 64 player BF4 for example you are still CPU limited on a 3770k @ 4.6 at ultra without AA. so you still see some improvement. so if you have a 144hz monitor it can still be worth it. Turning on MSAA is enough to switch the bottleneck back to the GPU almost completely(4x msaa will cut your mantle frame rate in half!) and you only see something like a 10% gain over DX11.

 

[itsmydamnation]

Are you serious? Other reviews have definitely shown that Mantle's benefits are far lesser on high end CPUs than on low end CPUs; including Anandtech.

I think the only time that changes is with CF or lower settings which makes the game more CPU bound. But anyone that's going to buy a powerful gaming CPU like a Core i7 likely isn't going to be playing at low or medium settings..

Did any reviews use CF?

care to link other review that match pclabs for relative performance difference between

edit: all are on ultra.

mantel

pclabs.pl
4770k 81.5
4770k@4.5 84.5
4670k 80.5
4670k@4.5 84.1

purepc.pl
4670k 108

golem.de
3770K 112

pcgameeshardware.de
4770k 128

DX11

pclabs.pl
4770k 64.5
4770k@4.5 69.8
4670k 61.3
4670k@4.5 69.2

purepc.pl
4670k 79

golem.de
3770K 63.8

pcgameeshardware.de
4770k 67

on relative performance improvements across review pcplab.pl scale at high clocks/core in multiplayer the least!

difference

pclabs.pl

81.5 64.5 1.263566
84.5 69.8 1.210602
80.5 61.3 1.313214
84.1 69.2 1.215318

purepc.pl
108 79 1.367089

golem.de
112 63.8 1.755486
pcgameeshardware.de
128 67 1.910448

 

[Venomous]

I don't feel like reading nearly 300 pages of posts.

Did this happen or no?

http://attackofthefanboy.com/news/radeon-r9-290x-mantle-api-ridicule-titan-gpu-bf4/

Far from it...

 

[wand3r3r]

I don't feel like reading nearly 300 pages of posts.

Did this happen or no?

http://attackofthefanboy.com/news/radeon-r9-290x-mantle-api-ridicule-titan-gpu-bf4/

There have been some pretty crazy gains, but for certain combinations.
The more CPU limited you are, the more you gain. For a single high end card with a high end CPU it's a lot less (perhaps closer to driver update gains). They claimed 290x crossfire gets up to 58% gains.

It's hard to decipher the results as a whole. A lot of posters seem to push the review which coincides with their views. Keep in mind this whole release is like a first version beta.


Where are the crossfire results?

I really want to see if the 290x crossfire has gone up, and how far.

If crossfire has such an advantage from mantle I think it will benefit every CPU.
The claim was amazing, but is it valid?

 

[bystander36]

Where are the crossfire results?

I really want to see if the 290x crossfire has gone up, and how far.

If crossfire has such an advantage from mantle I think it will benefit every CPU.
The claim was amazing, but is it valid?

Mantle's ability to reduce CPU bottlenecks alone would be enough to improve CF, but it will be nice to see more benchmarks. The few that I have seen came with the disclaimer that CF is buggy atm, which is likely why we haven't seen many.

 

[wand3r3r]

Mantle's ability to reduce CPU bottlenecks alone would be enough to improve CF, but it will be nice to see more benchmarks. The few that I have seen came with the disclaimer that CF is buggy atm, which is likely why we haven't seen many.

I tried to upgrade my MB, which turned out to be a dud or somehow borked. I have crossfire 290x but the 14.1 drivers are bad for mining so I reverted them. I have to reinstall BF4 anyway since I erased windows for the new MB which didn't work. Long story short, I cannot test for a while, and am in no hurry atm.

While I was playing BF4 in crossfire the usage was in the 50-100% range on a 4770k (at only 1200p). I guess that if they can get the usage up to the 100% range it will automatically have a huge effect, and there may be some other gains even if operating at full capacity. Whether this plays out remains to be seen, so far I've only seen AMDs (I believe, or was it EAs?) claim which is essentially marketing and no hard data from a major third party.

 

[3DVagabond]

It was Dice, and it was on their latest Dev build, not the public build. So, it might or might not show substantial gains with the 14.1 drivers and the current game patch.

 

[KompuKare]

Hardwareluxx.de /.com did do CF results:

http://www.hardwareluxx.com/index.p...tx-update-multiplayer-benchmarks.html?start=1

SB-E plus 2 x 290X. The next page has some frame time graphs. And Mantle looks a lot smoother except for those Mantle specific CPU spikes. Yes, 14.1 is beta or maybe alpha would be more appropriate. I guess some of these issues AMD hadn't seen in their internal testing but the ones they knew about, they should have fixed first.

Silly people for making 'promises', will they never learn? One of the first things Rory did was promise 28nm silicon shipping in 2012 and that promise probably had a lot with 7970 being rushed at in Dec 2012 without proper drivers and without being tuned for clocks/voltages. Which left people with the impression that GK104 was a lot more efficient than Tahiti when their perf/watt was actually fairly close. Not that I would encourage them to bin golden chips, but holding onto leaky Tahitis for a 7930 (like the 7870XT) would have been a good strategy.

 

[cytg111]

There have been some pretty crazy gains, but for certain combinations.
The more CPU limited you are, the more you gain. For a single high end card with a high end CPU it's a lot less (perhaps closer to driver update gains). They claimed 290x crossfire gets up to 58% gains.

It's hard to decipher the results as a whole. A lot of posters seem to push the review which coincides with their views. Keep in mind this whole release is like a first version beta.


Where are the crossfire results?

I really want to see if the 290x crossfire has gone up, and how far.

If crossfire has such an advantage from mantle I think it will benefit every CPU.
The claim was amazing, but is it valid?

This pretty much sums up what I've gathered so far.

As pr. the part i've bolded, this may well be the most active mantle thread on the whole of the interweb but it is problary also the most polluted. A shame how a pure numbers game can turn into something entirely different.

I'd be very much interrested in seeing CF results too, but not only CF, tri and quad as well. There is reports of upwards 100% gain on old running hardware such as core2duo's. In principle we should be able to push that CPU / GPU ratio into the highend as well, hex core ivey running tri or quad fire.
To understand the nature of a phenomenon, as a good scientist, you should push the extremes, where do we gain the most and where do we gain the least. We've got the "least" part down, and its close to 0, but I dont think we've seen the high end yet.

Speaking of, where is the Anand review? Not preview, review? Is it because its a beta product or are they just taking their sweet time and getting it right?

 

[itsmydamnation]

ryan said on B3d that anand review would be SP based. that was a few days ago who knows if it has changed ( thus pushed the time to release out). Personally i would love a methodical multiplayer review and included GPU OC's.

that one thing i find annoying in alot of these reviews...... 4.8ghz CPU and stock GPU's........

 

[cytg111]

ryan said on B3d that anand review would be SP based. that was a few days ago who knows if it has changed ( thus pushed the time to release out). Personally i would love a methodical multiplayer review and included GPU OC's.

that one thing i find annoying in alot of these reviews...... 4.8ghz CPU and stock GPU's........

I'd like them to keep the 4.8GHz clock on the cpu and then one stock GPU (290), then two, three and then four.
Yea, and then OC the cards in the end too.
Relevant data.

 

[3DVagabond]

The biggest issue with O/C'd reviews is the silicon lottery. The results could be quite worthless if the card the site has is an outlier.

 

[Carfax83]

care to link other review that match pclabs for relative performance difference between

edit: all are on ultra.

mantel

pclabs.pl
4770k 81.5
4770k@4.5 84.5
4670k 80.5
4670k@4.5 84.1

purepc.pl
4670k 108

golem.de
3770K 112

pcgameeshardware.de
4770k 128

DX11

pclabs.pl
4770k 64.5
4770k@4.5 69.8
4670k 61.3
4670k@4.5 69.2

purepc.pl
4670k 79

golem.de
3770K 63.8

pcgameeshardware.de
4770k 67

on relative performance improvements across review pcplab.pl scale at high clocks/core in multiplayer the least!

difference

pclabs.pl

81.5 64.5 1.263566
84.5 69.8 1.210602
80.5 61.3 1.313214
84.1 69.2 1.215318

purepc.pl
108 79 1.367089

golem.de
112 63.8 1.755486
pcgameeshardware.de
128 67 1.910448

That's a lot of data to go through, and tons of variables.....too many in fact to really come to any sort of valid conclusion.

Hardware and software configuration, the server player count and the campaign setting...all of these things matter.

 

[VulgarDisplay]

The crossfire numbers give a preview of mantle for high end 20nm GPUs. That has me far more excited for mantle moving forward. 50% faster likely for single cards on high end CPUs and maybe 80-100% faster in crossfire next year. I'm purely speculating, but the current benchmarks suggest this may be what we will be seeing on 20nm GPUs.

 

[kane15]

I don't feel like reading nearly 300 pages of posts.

Did this happen or no?

http://attackofthefanboy.com/news/radeon-r9-290x-mantle-api-ridicule-titan-gpu-bf4/

i know this has been replied to 10 times already but i can say it from first a hand experience i have a i7 920 OC to 3.2 and a 290x card so i am CPU bottlenecked and i play a ton of BF4 online, with DX id get dips into the 60's some times even the 50s but with mantle i never dip under 120 fps and i can usually stay at 144fps (have my fps capped at 144 to match my 144hz monitor) so for me no nvidia card on the planet would give me this type of performance, unless i built a new PC from the ground up around a new top of the line i7

 

[ASM-coder]

i know this has been replied to 10 times already but i can say it from first a hand experience i have a i7 920 OC to 3.2 and a 290x card so i am CPU bottlenecked and i play a ton of BF4 online, with DX id get dips into the 60's some times even the 50s but with mantle i never dip under 120 fps and i can usually stay at 144fps (have my fps capped at 144 to match my 144hz monitor) so for me no nvidia card on the planet would give me this type of performance, unless i built a new PC from the ground up around a new top of the line i7

Have you resolved your driver issues, after shut down and restart?

 

[kane15]

Have you resolved your driver issues, after shut down and restart?

yeah someone over on the AMD board said to restart the PC right after you do a cold boot up... i have no idea why this works but it does lol

 

[el etro]

The biggest issue with O/C'd reviews is the silicon lottery. The results could be quite worthless if the card the site has is an outlier.

Max core OC(with overvolting) is in most reviews past 1100Mhz. Most sites achieving poor overclocks are not overvolting the cards.

 

[Phynaz]

i know this has been replied to 10 times already but i can say it from first a hand experience i have a i7 920 OC to 3.2 and a 290x card so i am CPU bottlenecked and i play a ton of BF4 online, with DX id get dips into the 60's some times even the 50s but with mantle i never dip under 120 fps and i can usually stay at 144fps (have my fps capped at 144 to match my 144hz monitor) so for me no nvidia card on the planet would give me this type of performance, unless i built a new PC from the ground up around a new top of the line i7

100%+ increase...pics or it didn't happen.

 

[kane15]

100%+ increase...pics or it didn't happen.

http://oi57.tinypic.com/2nlh251.jpg http://oi60.tinypic.com/2hnqddj.jpg i took these awhile back, for some reason mantle looks brighter but if you adjust the brightness settings they look the same but i didnt want to play with any settings to show the compassion this is only on the test range and you can see i gained over 60fps and my graph is like a straight line, whats more impressive is actual MP when i would get those super low dips in fps those are now a thing of the past