Watchdogs2 benchmarks

[Head1985]

I don't get the hate for GTX1070 when it costs $350-370 and yet a reference GTX1080 that flew off the shelves and was defended and hyped to the moon on this forum is only 11% faster than an AIB 1070!

Hardware.fr shows 1080 FE beating 1070 AIB by only 12%. I am trying to understand how a $350-370 1070 is a failure but a $650-700 1080 FE that sold like hot cakes for months is constantly being overhyped. I keep reading on this very forum how GTX1080 beats GTX1070 by 30-35% in games but it's nowhere close to that. AIB 1080 is only about 19-23% faster than an AIB 1070.

Depends on game.That review is very old.New games have 25-33% gap GTX1070 vs GTX1080 in 1440p and 4k.
1440P:

Spoiler: New games benchmarks

Deus ex 27%

Dishonored2 25%

Mafia3 28%

BF1 27%

For some reason the gap is much more larger in new games than in old games.

 

[AtenRa]

You went that far from the benches from one game?

One game ??

Watch Dogs 2

Civilization VI

Battlefield 1

Gears Of War 4

There shouldn't be any $320 CPU from 2011 that could outperform a current latest generation $220 CPU in any 2016 game.
That only happens because of the luck of CPU progression at the $200-300 price segment. We are stack at 4C 8T in the $300-350 price point for more than 7 years now (since Core i7 920 back in 2009).

Look at the graphs above, except of BF1 which is GPU limited with the high-end CPUs, there is less than 15-20% performance increase between the Core i7 2600K and the 6700K. And 7700K next month will launch at the same $350 price point and will bring NOTHING over the 6700K for another year. This is pathetic really, we could have 6-8 core Skylake next year at the same price points easily. Instead, we get mobile SKUs in desktop the last 3-4 years, the last real Desktop mainstream SKUs where the Core i7 920 and 2600K, after that we get mobile SKUs for the same price/perf segment of 2011.

 

[tg2708]

@Face2Face yea I know it's the better overall card but I just like the 980ti a bit more barring mine sucks at overclocking. Also does. Also does boost 2.0 let the card consume more power if the card boosts past it's advertised clocks? Also trying to mod by bios would probably leave me with a broken card lol.

 

[Head1985]

One game ??

Watch Dogs 2

Civilization VI

Battlefield 1

Gears Of War 4

There shouldn't be any $320 CPU from 2011 that could outperform a current latest generation $220 CPU in any 2016 game.
That only happens because of the luck of CPU progression at the $200-300 price segment. We are stack at 4C 8T in the $300-350 price point for more than 7 years now (since Core i7 920 back in 2009).

Look at the graphs above, except of BF1 which is GPU limited with the high-end CPUs, there is less than 15-20% performance increase between the Core i7 2600K and the 6700K. And 7700K next month will launch at the same $350 price point and will bring NOTHING over the 6700K for another year. This is pathetic really, we could have 6-8 core Skylake next year at the same price points easily. Instead, we get mobile SKUs in desktop the last 3-4 years, the last real Desktop mainstream SKUs where the Core i7 920 and 2600K, after that we get mobile SKUs for the same price/perf segment of 2011.

Gamegpu CPU tests are bad.I am 100% sure they are testing in worng areas that are not cpu heavy/bottleneck.
look at this

And then real life testing here:
http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...astered-gpu-perf.2491354/page-3#post-38603452

 

[Innokentij]

Well it boosts 1430 and settle at 1405 once I raise the temp limit to 91 and power target to 115 and no I don't have a friend with a similar system to test it in. Cards always boost over guaranteed clocks though.

Set it to 130% powertarget.

 

[tg2708]

Set it to 130% powertarget.

Only goes up to 115 percent

 

[AtenRa]

.I am 100% sure they are testing in worng areas that are not cpu heavy/bottleneck.

Well the Downtown benchmark they did in Watch Dogs 2 seems to be a heavy CPU area

vs

 

[Innokentij]

Only goes up to 115 percent

Get evga precision x 5.3.11

 

[Majcric]

Well the Downtown benchmark they did in Watch Dogs 2 seems to be a heavy CPU area

vs

I've been CPU testing that same strip and found CPU drops to be the most at the end heading off into the forest area. Right at the end of that long runway CPU utilization will drop and the fps will follow.

 

[Majcric]

DP

 

[Carfax83]

Problem with pascal is that pascal dont scalling with frequency.I tested it with MY GTX1070.I tested 1500-2150Mhz and card stopped scalling at 1700-1800mhz.You can add whatever you want, but performance dont change much after that(its not in all games, but in most games gtx1070 dont get much from oc if its already running at 1924mhz)

This might be specific to the GTX 1070, and likely due to it being bandwidth limited vs the GTX 1080 and Titan X. GTX 1070 only has 256GB/s bandwidth, which really isn't that much for a high end card with today's games. Of course Pascal's awesome compression maximizes the bandwidth and then some, but it will eventually run into a wall no matter what..

 

[Carfax83]

PCgameshardware.de updated their benchmarks with CPU overhead measuring, and this more than anything explains the gulf between AMD and NVidia in Watch Dogs 2, much like the original game.

DX12 would do a world of good for this game, because it really seems to be CPU intensive. That said, is the game any good?

I kind of wanted to buy it, then I held off as I never finished the original..

 

[Head1985]

This might be specific to the GTX 1070, and likely due to it being bandwidth limited vs the GTX 1080 and Titan X. GTX 1070 only has 256GB/s bandwidth, which really isn't that much for a high end card with today's games. Of course Pascal's awesome compression maximizes the bandwidth and then some, but it will eventually run into a wall no matter what..

Not really i have memory at 9500mhz=304Gb/s.I think its lack of rasterizer/GPC= low pixel fillrate.
Its only thing where its slower than even stock GTX980TI(1100mhz)

 

[Face2Face]

@Face2Face yea I know it's the better overall card but I just like the 980ti a bit more barring mine sucks at overclocking. Also does. Also does boost 2.0 let the card consume more power if the card boosts past it's advertised clocks? Also trying to mod by bios would probably leave me with a broken card lol.

Are you increasing the voltage at all? If not, that could be your issue.

Well the Downtown benchmark they did in Watch Dogs 2 seems to be a heavy CPU area

What a great showing for the AMD FX series. Looks like the MOAR CORES meme may be eating crow soon enough.

 

[dogen1]

Not really i have memory at 9500mhz=304Gb/s.I think its lack of rasterizer/GPC= low pixel fillrate.
Its only thing where its slower than even stock GTX980TI(1100mhz)

But overclocking should increase the pixel and texel fill rate at the same rate as flops.

 

[Head1985]

But overclocking should increase the pixel and texel fill rate at the same rate as flops.

yeah but there must be some bottleneck.They cut out that GPC for reason=to create bottleneck and to gtx1070 not be abble match gtx1080 performance.
GTX1070 at 1800-1900mhz have less pixel fillrate than 980TI at 1000-1100mhz.If you oc 1070 to 2150Mhz you still dont get much more because its already running at 1800-1900Mhz.

 

[Carfax83]

Not really i have memory at 9500mhz=304Gb/s.I think its lack of rasterizer/GPC= low pixel fillrate.
Its only thing where its slower than even stock GTX980TI(1100mhz)

I think you need to take into account the fact that the GTX 1070 is already running at a very high clock speed. PC gamer did some overclocking in their 1070 review. They managed to get an extra 15% above stock from overclocking the GTX 1070, both core and VRAM. The thing is, the GTX 1070 is obviously more limited than the GTX 1080, especially in terms of bandwidth. Also, there is diminishing returns. Every GPU has a sweet spot where you can gain the most performance, and if you try to exceed that threshold, the performance returns are dramatically diminished..

For the 980 Ti I think it was probably about 1400-1440. With the GTX 1080, it's 2000, and for the GTX 1070, it's probably 1900 or so.. This doesn't necessarily mean you run into a wall, just that the gains beyond those ranges falls off dramatically due to architectural or bandwidth limitations..

 

[LTC8K6]

Look at the graphs above, except of BF1 which is GPU limited with the high-end CPUs, there is less than 15-20% performance increase between the Core i7 2600K and the 6700K. And 7700K next month will launch at the same $350 price point and will bring NOTHING over the 6700K for another year. This is pathetic really, we could have 6-8 core Skylake next year at the same price points easily. Instead, we get mobile SKUs in desktop the last 3-4 years, the last real Desktop mainstream SKUs where the Core i7 920 and 2600K, after that we get mobile SKUs for the same price/perf segment of 2011.

There is no 6700K in those charts, nor is there a 6600K.

The used a 6700, which is considerably slower than a 6700K, and a 6600, which is noticeably slower than a 6600K.

The 7700K at $350 will be decently ahead of the 6700K, and well ahead of the 6700 that was used in those benches.

The 7600K will also be well ahead of the 6600 used in the benches.

 

[ShintaiDK]

http://www.pcgameshardware.de/Watch-Dogs-2-Spiel-55550/Specials/Test-Review-Benchmark-1214553/

Shame if you got the graphics power but not the CPU.

 

[ShintaiDK]

Do you have that backward? 1060 3g is the only card that shows massive perf loss when upping quality. Fury gains a single fps.

Not at all.

 

[VirtualLarry]

Every i5 can hit 4.5+

Uhh, NO, Happy. Just NO. At least, not for the "gimp bin" i5-6400. Mine wasn't stable at 4.5Ghz, so I had to settle for 4.455. Not going to push higher than 1.410V, and 85C on air.

 

[casiofx]

Uhh, NO, Happy. Just NO. At least, not for the "gimp bin" i5-6400. Mine wasn't stable at 4.5Ghz, so I had to settle for 4.455. Not going to push higher than 1.410V, and 85C on air.

What cooler did you use?

 

[Erenhardt]

Not at all.

Or maybe yes?

 

[casiofx]

Or maybe yes?

I don't think GTX1060 3GB owners will be dumb enough to use High Resolution Texture Packs which your source uses.

 

[poofyhairguy]

One game ??

Watch Dogs 2

Civilization VI

Battlefield 1

Gears Of War 4

Wow. I can't help but to look at those benches and admit those saying the FXs would eventually catch up with Intel on gaming were right.

Good news for Zen.