Any hd7870 AND hd7970 card owners want to help w/ experiment?

[tviceman]

I need two people who have fast CPU's (most preferably Intel quadcores or better with decent overclocks) who are running an hd7870 and hd7970 AND own Sleeping Dogs.

The HD7970 owner
I'd like the owner of the hd7970 to run the game's built-in benchmark twice at 1080p with all in-game settings maxed out including anti-aliasing and the high resolution texture pack, vsync disabled of course, with their card running at two different speeds: 1050mhz on the core and 1500mhz on the vram (6000mhz effective on the vram) for the first run, and 1050mhz core but only 1000mhz (4000mhz effective) on the vram on the second run (thus limiting the memory bandwidth to 192 gb/s).

The hd7870 owner
I'd like the owner of the hd7870 to run the game's built-in benchmark once at 1080p with all the in-game settings maxed out including anti-aliasing and the high resolution texture pack, vsync disabled, with their card running at stock core speeds and 1500mhz (6000mhz effective) on the vram.



Please help me and please post screenshots of the run results, as well as afterburner read outs after the run to verify all the core/mem speeds. I would greatly, greatly appreciate it. I want to see what effect memory bandwidth has if the hd7970's and hd7870's bandwidth's are both equal to the gtx670/680. Not trying to start a war about anything, this is just for curiosity's sake.

 

[VulgarDisplay]

Buy me the game and I'm in.

 

[Dravonic]

I have an 8350 @ stock. If that's good enough, I'm willing to provide the 7970 data.

 

[tviceman]

Buy me the game and I'm in.

Do you have a hd7870?

 

[tviceman]

I have an 8350 @ stock. If that's good enough, I'm willing to provide the 7970 data.

I'd say yes, but I have an i7-2600k running @ 4.4ghz, and I just don't want that kind of discrepancy to muddy the results. Thanks though.

 

[VulgarDisplay]

Do you have a hd7870?

7970 and a 2500k at 4.5ghz

 

[tviceman]

7970 and a 2500k at 4.5ghz

Bah! Plenty of 7970 owners here!

 

[VulgarDisplay]

Bah! Plenty of 7970 owners here!

Buy me a 7870 and Sleeping Dogs and I'm in. :biggrin:

 

[lavaheadache]

7970 Lightning and 7870 Hawk here... Sleeping Dogs, negative

 

[footballrunner800]

Im downloading sleeping dogs right now to aid in this post!

 

[footballrunner800]

3930k is at 4.6

 

[chimaxi83]

I can do this for ya if you need more results, it'd have to be tomorrow though

 

[MustangSVT]

why do you need screenshots if its only for your "curiosity" ?

hmmm....

 

[Termie]

I actually did this test using 3dMark Fire Strike (which may favor AMD a bit more than most benchmarks).

I found that at the same clock speeds, my GTX670 is about 9% faster than my HD7870 (clocks - 1175/5800). But the 670 has far more potential memory bandwidth, as the HD7870's memory gives up where a stock 670 starts.

And you should just use a free benchmark if you want additional data points on this. Unigine Heaven 4.0 would be a good one.

 

[Dravonic]

I'd say yes, but I have an i7-2600k running @ 4.4ghz, and I just don't want that kind of discrepancy to muddy the results. Thanks though.

Well, I got curious so I did it anyway. Might as well post the results.

1050/1500:

1050/1000:

 

[footballrunner800]

1050/1500:

1050/1000:

 

[Bryf50]

7870s are almost all 4800mhz stock. You're not getting 6000mhz out of a 7870s vram. Best I can do on mine is 5400mhz.

 

[Midwayman]

i7-2600k and a 7870, but no sleeping dogs. Oh well.

 

[Shephard]

I got 3570k and 7870 myst if you want to buy me sleeping dog.

 

[BoFox]

Well, I got curious so I did it anyway. Might as well post the results.

1050/1500:

1050/1000:

That's 15% difference due to bandwidth.

That translates to 3% performance difference for every 10% increase in bandwidth.

 

[BoFox]

I need two people who have fast CPU's (most preferably Intel quadcores or better with decent overclocks) who are running an hd7870 and hd7970 AND own Sleeping Dogs.

The HD7970 owner
I'd like the owner of the hd7970 to run the game's built-in benchmark twice at 1080p with all in-game settings maxed out including anti-aliasing and the high resolution texture pack, vsync disabled of course, with their card running at two different speeds: 1050mhz on the core and 1500mhz on the vram (6000mhz effective on the vram) for the first run, and 1050mhz core but only 1000mhz (4000mhz effective) on the vram on the second run (thus limiting the memory bandwidth to 192 gb/s).

The hd7870 owner
I'd like the owner of the hd7870 to run the game's built-in benchmark once at 1080p with all the in-game settings maxed out including anti-aliasing and the high resolution texture pack, vsync disabled, with their card running at stock core speeds and 1500mhz (6000mhz effective) on the vram.



Please help me and please post screenshots of the run results, as well as afterburner read outs after the run to verify all the core/mem speeds. I would greatly, greatly appreciate it. I want to see what effect memory bandwidth has if the hd7970's and hd7870's bandwidth's are both equal to the gtx670/680. Not trying to start a war about anything, this is just for curiosity's sake.

Reducing 7970's bandwidth to the same bandwidth "count" as GTX 680 will make the 7970 behave like as if it's more bandwidth-restrained - due to limited crossbar access to the RAM chips. It could actually be as bad as 192-bit, vs GTX 680's 256-bit....

Nvidia pretty much designed their GPUs with better bandwidth utilization - slightly more efficiently utilized with the majority of games overall. It is implied by one of Futuremark's benches that pretty much shows the effect of bandwidth so well.

Anyway, it looks like 7970GE would still hover at around 220VP, after reduced bandwidth hurt its performance by 15%.

Just for fun, let's take this game as an average game, and assume that GTX 680 is rated at 235VP for this.
235VP is 7% faster than 220VP.

What if 7970GE's bandwidth was utilized 33% better without limited crossbars, etc.? 33% more bandwidth = 10% more performance. Then it'd be closer to being as efficient specs-wise as Pitcairn 7870's, churning out 242VP with a straight 256-bit bus at 6000MHz.

The 384-bit bus does not help 7950 much at all vs 7870, with the former having 12% more GFLOPs rating, yet only 11.5% more VP (despite having 56% more bandwidth). That is another indication that the pure Hz speed of the GPU is overall a bit more effective than increasing the number of shaders. Another thing is that the driver program for Tahiti is a bit different than for Pitcairn. It's like AMD shifting gears to 6th gear for Tahiti, from 5th for Pitcairn. It affects everything, including power consumption ratio, etc..

That's just a shot in the barrel talk.

 

[Termie]

Also, I should note that both my 670 and 7870 exhibit 22% scaling with memory overclocking. It seems from the 7970 benches above that the 7970 scales at 30%, but that is an incorrect conclusion. That's with extreme bandwidth constraint. If you instead overclock the memory, you'll find that its memory scaling is actually lower than the the 7870 or 670/680, because it is fundamentally less bandwidth constrained due to its wider bus. I've seen data from other users that demonstrated approximately 18-20% memory scaling on the HD7900 series.

Honestly if you want to run this test, we need to be a bit more scientific about the approach. I'm really not sure what the OP's hypothesis was and whether the data tend to support it or refute it.

 

[tviceman]

Thanks for the benchmarks with the he7970's. I didn't realize the 7870's vram chips were clocked so low to begin with. Regardless. Anyone with 7870's and this game, it would still interest me to see be benchmarks with stock core speeds and max vram oc'd speeds.

 

[tviceman]

Honestly if you want to run this test, we need to be a bit more scientific about the approach. I'm really not sure what the OP's hypothesis was and whether the data tend to support it or refute it.

I want to see how the hd7970 and hd7870 perform against a gtx670/680 in this game with the highest anti-aliasing settings when all the cards have similar memory bandwidth. Some people argue that GK104's compute abilities suck, but whether or not true I don't think that is the case of what is causing the larger performance drop with Kepler (vs. Tahiti) when maxing out the AA options in this game. I think it's mostly memory bandwidth related (a noticeable area of weakness that pretty much the entire Kepler lineup suffers from).


Soooo, for the sake of curiosity and discussion of memory bandwidth / compute functionality and this game, could you run the game's benchmark with your hd7870 at stock core clocks and maxed out memory overclocks? It appears your 7870 can get to at least 1400mhz vram, which is 179.2 gb/s bandwidth, 93.3% the bandwidth of an hd7970 @ 1000mhz vram or a gtx670/680 @ 1500mhz.

 

[tviceman]

If anyone else with an hd7870 and also has this game is willing to throw a benchmark in with their core set to AMD's reference speeds and the memory overclocked maxed out, I'd appreciate it!