Upgraded from 670 to 780 to 980 with same specs

[MissChief]

Well, almost. After having problems with my 780 (6GB) I've now got a 980 as it's replacement. Other than an SSD nothing else has changed in my system. The only exception is the overclock I used to have on my core i5 2500k, 4.2GHz was removed during testing of the 780 after 5 beeps indicated CPU but it was actually the card. I've now reapplied the overclock but I thought it would be interesting to see what effect the graphics upgrades have had on the performance of my system. Unfortunately I don't have any other Benchmarks saved other than Unigine's Valley but here we go anyway:

670

780

980 (without CPU OC)

As you can see, not a huge difference between the 780 and the 980 which surprises me. I'm off to reapply my OC and see if that makes a difference.

OK, 4.2GHz OC to the CPU applied, confirmed in CPU-Z prior to the run.

Weirdly the results are actually LOWER. Not sure what happened on this one, the figures seem WAY off so I ran it again.

My second run which seems much more like it:

 

[NeoVo]

I would say the benchmark is CPU bottlenecked, re apply your over clock on the 2500k and re-run the benchmark.

 

[NoNameNoNumber]

I doubt the 2500k is bottle necking a single 980

 

[eclap]

I doubt the 2500k is bottle necking a single 980

In benchmarks like this, a difference between stock 2500k and overclocked 2500k will show a difference. Similar in 3Dmark, the CPU will affect the physics and combine score, also the overall P Score. Not the graphics score though, that relies almost purely on the GPU.

 

[Rroff]

I don't recall valley being significantly CPU bottlenecked - though the Maxwell GPUs seem to respond a bit to i7 4+4 and 6+6 CPUs for some reason (not sure if its some driver thing).

A decent boosting 780 can give a stock 980 a good run mind - not sure how your 780 and 980 stack up compared to reference clocks.

 

[MissChief]

Old card had a 1019 MHz base and will boost up to 1071 MHz and 6 GB of memory clocked at the standard 6.0 GHz (3GHz x2) whereas the new GTX 980 SOC Has 1228MHz core with 1329MHz boost. Memory is 7010MHz (3.505Ghz)

 

[setter]

Iirc valley and heaven are pretty much gpu based, I've ran it with 3570k, 3770k, 4770k and 4790k. All with the same cards, single 670, sli 670, single 780, sli 780 and single 980ti. Not much difference tbh. Various overclocks on theese gpu's.

 

[eclap]

Iirc valley and heaven are pretty much gpu based, I've ran it with 3570k, 3770k, 4770k and 4790k. All with the same cards, single 670, sli 670, single 780, sli 780 and single 980ti. Not much difference tbh. Various overclocks on theese gpu's.

Sure, they're gpu based. CPU will still make a slight bit of a difference in Unigine benches.

980 should still score more than a 780, by quite a bit. OP, can you run FireStrike and post your score? That will give us a bit more of an insight to whether you have an issue or not.

 

[Rroff]

Old card had a 1019 MHz base and will boost up to 1071 MHz and 6 GB of memory clocked at the standard 6.0 GHz (3GHz x2) whereas the new GTX 980 SOC Has 1228MHz core with 1329MHz boost. Memory is 7010MHz (3.505Ghz)

From a quick look at the leaderboard the 780s do quite well in Valley - obviously they can't be directly compared but clock v clock a 780 scores the same as a 980 that is running a ~14% higher boost clock i.e. there are 2 people with the same score one on a 14xxMHz 780 the other on a 16xxMHz 980.

 

[MissChief]

Firestrike: http://www.3dmark.com/3dm/8686980

 

[SouthEastBlue]

780 scores stupidly high in both Heaven and Valley in comparison to GTX 970 and 980 as far as I've seen, this isn't the same in games though.

 

[Joe!]

Valley isn't necessarily a good indicator of performance, especially at 1080p. It's really CPU bound. I personally went from an i5-3450 to an i7-5820k and went from the lowest 980Ti to the highest 980Ti on Valley@1080p - The difference a good CPU makes. I can't say I was CPU bottlenecked in games whatsoever though.

OT (somewhat): I am pleased you managed to get a GTX 980, I suppose the whole 3.5gb VRAM thing worked in your favour.

 

[Marine-RX179]

I say it's probably to do with the fact that the settings used for Heaven, and how the bench is, with Extreme Tessellation and 8xAA being used at the same time (which wouldn't really be used in real world gaming environment).

The 980 comparing to the 780 has a narrower bus-size and lower memory bandwidth and far fewer texture units, considering the 980 is a "skinny" GPU rather than a "full fat" GPU like the 780 is.
http://www.bit-tech.net/hardware/graphics/2014/09/19/nvidia-geforce-gtx-980-review/1

I think the 8xAA is probably dragging down the performance of the 980 the most, and if it was lowered to say 4xAA, the performance grab between the 980 and the 780 should widen I think.