AMD Radeon R9 300 series 3DMARK 11 score

[Gloomy]

It's going to be interesting to see how AMD's brute force HBM stacks up against big Maxwell's bandwidth efficiency improvements.. The enhanced color compression of Maxwell adds about 25% to the actual effective bandwidth, so GM200's bandwidth should be close to 400 GB/s. Also if NVidia doubles the L2 cache from 2MB to 4MB, then bandwidth efficiency will be even greater. I don't expect big Maxwell to have issues with 4K resolution considering that GM204 already competes with Hawaii in that aspect very well, even though Hawaii has significantly higher raw bandwidth.

But if R9 390x has nearly twice as much bandwidth as GM200, then it should conceivably handle 4K plus additional eye candy effects better than it's competitor.

Tonga already showcased Maxwell-level memory compression. AMD actually claims 40% and this is evidenced in tests:

With AMD maintaining their same texture units and same tex:FP32 ratio for Tonga, the end result is that R9 285’s texturing performance is virtually identical to R9 280’s. Ignoring any possibility of caching or bandwidth bottlenecking for the moment, R9 285 can push texels just as well as R9 280 could. [with 30% less bandwith]

Unlike our texel test, AMD’s delta color compression technology introduced on GCN 1.2 has an incredible impact on R9 285’s pixel throughput. This pixel test is normally memory bandwidth bound, providing something that approaches a best case scenario for AMD’s compression technology. As a result despite possessing nearly 30% less memory bandwidth than the R9 280, the R9 285 tops our charts at 19.9 GPix/sec, blowing past the R9 280 by 68%. Even the R9 290 with its 512-bit memory bus and doubled ROP count still falls short here by over 3GPix/sec, or 16%.

This benchmark in a nutshell is why AMD can deliver the average performance of the Tahiti based R9 280 without Tahiti’s memory bandwidth. By improving their color compression to this point AMD can significantly reduce their memory bandwidth requirements on GCN 1.2, allowing them to do more with less. In real games the result won’t be anywhere near this remarkable since this is a pure pixel fillrate test, but it goes to show that AMD has been able to neutralize their memory bandwidth deficit in graphics workloads.

70% is the best case for it. 40% is an optimistic expectation. But it is typical to see the 285 match the 280-- so in most cases it is at least a 30% efficiency improvement.

 

[Carfax83]

Tonga already showcased Maxwell-level memory compression. AMD actually claims 40% and this is evidenced in tests.

Yes I know that, but thats just one part of the equation. The other part is the increased size of the L2 cache, which further amplifies bandwidth efficiency. GM200 will likely have twice as much cache as GM204 if I had to guess..

Anyway, I'm not saying that GM200 is going to trounce AMD's 380x/390x at 4K and above resolution, but it will be competitive, which is all that matters.

4K won't be big for about another two years or so anyway.

 

[Gloomy]

Yes I know that, but thats just one part of the equation. The other part is the increased size of the L2 cache, which further amplifies bandwidth efficiency. GM200 will likely have twice as much cache as GM204 if I had to guess..

Anyway, I'm not saying that GM200 is going to trounce AMD's 380x/390x at 4K and above resolution, but it will be competitive, which is all that matters.

4K won't be big for about another two years or so anyway.

You are too optimistic. :|

If what we know now about both GPUs are true then AMD's flagship is going to be faster by at least 30%.

 

[Carfax83]

You are too optimistic. :|

If what we know now about both GPUs are true then AMD's flagship is going to be faster by at least 30%.

I could say you're being optimistic as well. Out of curiosity though, how do you get 30%?

 

[Gloomy]

I could say you're being optimistic as well. Out of curiosity though, how do you get 30%?

No math, just a random guess. If AMD puts double the amount of bandwidth on a GPU, it's likely that the GPU can use it, so it would have to be twice as fast. A 390X should be approaching double over a 290X. I didn't even account for the improvements AMD put in Tonga like the bandwidth efficiency, tessellation efficiency, etc.

Since GM200 is only a 50% increase in bandwidth over GM204, then that's the general ballpark where it will land. Again, not accounting for whatever efficiency improvements Nvidia puts in.

Of course nothing ever really works out simple like that so you have to have some fudge either way.

 

[lavaheadache]

I will say this.... Can't wait to sell my Titan Z to buy a 390X. :looks over at the Zed: "sorry to see you go but looks like launch time is a comin'"

 

[Carfax83]

No math, just a random guess. If AMD puts double the amount of bandwidth on a GPU, it's likely that the GPU can use it, so it would have to be twice as fast. A 390X should be approaching double over a 290X. I didn't even account for the improvements AMD put in Tonga like the bandwidth efficiency, tessellation efficiency, etc.

Since GM200 is only a 50% increase in bandwidth over GM204, then that's the general ballpark where it will land. Again, not accounting for whatever efficiency improvements Nvidia puts in.

Of course nothing ever really works out simple like that so you have to have some fudge either way.

Bandwidth is just one factor in the overall performance of a GPU. Shader array is more important, and Maxwell shaders are extremely efficient. A GTX 970 1600+ shader cores easily competes with, and even outcompetes a R9 290x which has 2800+ GCN cores for example, with a lot less bandwidth to boot..

While I admit that cross architectural comparisons like that are foolish, I'm saying that you just can't go off the numbers and assume.

I do think that the R9 390x is going to be a beast though. It's probably going to be AMD's best chance in years at decisively taking back the performance crown and keeping it for a while.

 

[Rvenger]

Thread cleaned and reopened - please keep it civil and on topic. No callouts or trolling please.


Rvenger
Anandtech Super Moderator

 

[RampantAndroid]

Doesn't 3D Mark upload the score and verify it? Is there an upload of this anywhere that might lead to some credence to it being real and not a photoshop?

 

[dmoney1980]

Let's hope that AMD gets it right with the stock cooler this time

 

[Plimogz]

[...]Can't wait to sell my Titan Z to buy a 390X.[...]

Heh, somehow I didn't think you sold off previous flagships when upgrading -- though it makes total sense, resale value and all that.

Still, if my suspicions hold true, it seems that you wouldn't so much be selling it as renting it out -- after all, it, or another much like it, will likely find its way back home after a while, no?

***

These 390x rumors are making me nervous about maybe regretting my 290s.

 

[dmoney1980]

These new cards - for both Nvidia and AMD- need to release in time for Witcher 3

 

[lavaheadache]

Heh, somehow I didn't think you sold off previous flagships when upgrading -- though it makes total sense, resale value and all that.

Still, if my suspicions hold true, it seems that you wouldn't so much be selling it as renting it out -- after all, it, or another much like it, will likely find its way back home after a while, no?

***

These 390x rumors are making me nervous about maybe regretting my 290s.

now that you know my strategy I must kill you, lol. I never hold on to cards while they are still hot. This is an expensive hobby you know. Considering my Z only owes me a grand I should do ok moving it.

 

[KaRLiToS]

Faker than fake boobs.

2 x R9 290 Stock clock, meh :thumbsdown:

http://www.3dmark.com/3dm11/9300727

 

[cmdrdredd]

Faker than fake boobs.

2 x R9 290 Stock clock, meh :thumbsdown:

http://www.3dmark.com/3dm11/9300727

LOL! Also I wondered, why the heck would they test 3dmark 11? It's not representative of a modern game engine.

 

[Techhog]

Faker than fake boobs.

2 x R9 290 Stock clock, meh :thumbsdown:

http://www.3dmark.com/3dm11/9300727

... Yep, that's the same result. Well, I guess it's time to leave this thread.

 

[KaRLiToS]

Actually, when the R9 290x released, Smoke, a LN2 overclocker did a run of 3dmark 11. This score was posted around the world... it was legit score.

http://rog.asus.com/268622013/overc...dmark-11-record-with-quad-r9-290x-crossfirex/
http://hwbot.org/submission/2443921_smoke_3dmark11___performance_radeon_r9_290x_22640_marks

Smoke Smashes 3DMark 11 Record With Quad R9 290X CrossFireX

 

[Rvenger]

I guess we don't need this thread anymore since the benchmarks are fake. Closed!


-Rvenger