Watchdogs2 benchmarks

[Bacon1]

And boy GCN pre Polaris looks bad with features upped on the PCGH review. Fiji went down fast!

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

 

[Bacon1]

Except GameGPU shows a 220W FX-9590 barely matching a 3-year old 84W i7-4770K at stock, but I guess throwing an expensive Intel chip + GPU bound scenario and/or scripted benchmark makes the comparison more interesting to some people.

If you took the chip off your shoulder you'd see I'd posted a similar graph as well, where they used the 1080 and showed there was a difference @ 1080p. Point is the game is GPU bound not CPU bound. You can get away with a cheaper CPU with many threads.

 

[AtenRa]

OK so first off, could you show an example of CrossFire working in a modern title but SLI not working in that same title? Again, the point I was making was that there are a lot of games that simply can't use AFR. Period. You have twisted the original point.

If the GTX 1060 is overpriced, then what does that make the RX 480, which offers roughly the same perf/$ ratio?

No, NVIDIA would love to be able to sell people on more GPUs, that's just straight up more money into their pockets. The reality is that games are increasingly using rendering techniques that make AFR very difficult.

It's a good thing that people won't be "tempted" to buy dual GTX 1050 Tis or GTX 1060s, when they could get a much better gaming experience (and in the case of SLI'd 1060s, for less money) on average by going with a GTX 1070.

I said lack of SLI support (GTX 1060/1050) and lackluster scaling for the rest (only up to 50% in most SLI supported games)

I said NVIDIA wants to sell smaller dies (300mm2 GP104) at higher margins ($600-700 GTX 1080) than selling 2x GP106 GTX 1060 (2x 200mm2 = 400mm2) at 2x $250 = $500

If 2x 1060 was as fast or faster than a single 1080, NVIDIA would loose GTX 1070 sales, clearly they dont want that. Im sure you of all people would understand this, its only matter of getting more money and nothing more. Consumer is left without SLI support at $250 and lower price point because NV wants to sell 300mm2 dies at $400 to $600-700.
Well thats fine if NVIDIA doesnt want the consumer to be able to get more performance for less, but thankfully NVIDIA is not the only dGPU company, AMD supports CF even at the low $100 price point. AMD also invest in mGPU with new technologies like Frame Pacing in DX-12, so that is another win for the consumer because we have more choices.

As the mGPU Watch Dogs 2 benchmarks shows, 2x RX 480 8GB at $420 today is as fast as $560 (cheapest today) GTX 1080. Im sure GTX 1060 SLI could also reach the same performance IF NVIDIA wanted.

 

[happy medium]

When was the last time we have seen 7 out of the top 10 slots in a review belong to a single video card maker? In this case Nvidia.

And look at the 780ti now in this review? Just behind the rx480.

 

[Erenhardt]

Not bad performance. I expected it to be a gimp showcase.

Except GameGPU shows a 220W FX-9590 barely matching a 3-year old 84W i7-4770K at stock, but I guess throwing an expensive Intel chip + GPU bound scenario and/or scripted benchmark makes the comparison more interesting to some people.

Your wording is kinda strange, maybe it is a language barrier. Just to sum your post up: almost 2 years older AMD stock CPU is equal to younger intel stock CPU but is rated at higher power consumption (probably because it is produced on much older manufacturing process)?

That seems nice for all those who have cheap Vishera CPUs. Its almost as fast as intel latest and greatest.

 

[Bacon1]

When was the last time we have seen 7 out of the top 10 slots in a review belong to a single video card maker? In this case Nvidia.

When you use 2 $1000+ cards from one maker and not the other? They could have thrown in a Fury Pro Duo.

 

[yellow_eggplant]

Not bad performance. I expected it to be a gimp showcase.



Your wording is kinda strange, maybe it is a language barrier. Just to sum your post up: almost 2 years older AMD stock CPU is equal to younger intel stock CPU but is rated at higher power consumption (probably because it is produced on much older manufacturing process)?

That seems nice for all those who have cheap Vishera CPUs. Its almost as fast as intel latest and greatest.

Also, the 4 year old 8350 handily beating a newer and more expensive i5 6600.

This game uses multiple threads well at least.

 

[Sweepr]

Just to sum your post up: almost 2 years older AMD stock CPU is equal to younger intel stock CPU but is rated at higher power consumption (probably because it is produced on much older manufacturing process)?

Sorry but the only one who didn't understand is you. A 2011 95W i7-2600K at stock comes very close to a 2014 220W FX-9590 - both are 32nm chips and the Intel chip has a huge overclocking headroom on top. So no, it doesn't look good for FX unless you manage to find a very light scene were any CPU would perform the same (hi Guru3D!).

 

[happy medium]

When you use 2 $1000+ cards from one maker and not the other? They could have thrown in a Fury Pro Duo.

right, I forgot about that card.

 

[AtenRa]

Sorry but the only one who didn't understand is you. A 2011 95W i7-2600K at stock comes very close to a 2014 220W FX-9590 - both are 32nm chips and the Intel chip has a huge overclocking headroom on top. So no, it doesn't look good for FX at all unless you manage to find a very light scene were any CPU would perform the same (hi Guru3D!).

Funny you missed the FX8350 at default being faster than the 3 year younger and more expensive Core i5 6600 14nm Skylake

 

[happy medium]

Also, the 4 year old 8350 handily beating a newer and more expensive i5 6600.

This game uses multiple threads well at least.

Yes , it seems like its using all available cores.

 

[Bacon1]

Sorry but the only one who didn't understand is you. A 2011 95W i7-2600K at stock comes very close to a 2014 220W FX-9590 - both are 32nm chips and the Intel chip has a huge overclocking headroom on top. So no, it doesn't look good for FX unless you manage to find a very light scene were any CPU would perform the same (hi Guru3D!).

Why do you keep using the 9590? Its nothing more than OC'd 8350 (2012 release). It also came out in June 2013 not 2014. The 8350 is faster than the i5 6600 which came out in 2015.

Also stop using TDP as actual wattage used when we have no testing showing how much wattage is actually used / required for the game. CPU graphs show they are clearly not at 100% so wouldn't be using full power.

 

[happy medium]

more cpu stuff. My i3 6100 might need a little overclock for this game.

 

[Sweepr]

Why do you keep using the 9590? Its nothing more than OC'd 8350. It also came out in June 2013. The 8350 is faster than the i5 6600

In this game, but it's slower in the majority of titles, and usually by larger margins than FX's advantage here. Not to mention an almost 6-year old i7-2600K would destroy any FX when both are overclocked, so I'm sorry but I don't see anything to brag about.

clearly not at 100% so wouldn't be using full power.

Wrong:

Ps: Your Guru3D chart where they tested a very light scene (hence why there's zero CPU scaling) doesn't count.

Back to the GPU discussion now, bye.

 

[Bacon1]

In this game, but it's slower in the majority of titles, and usually by larger margins than FX's advantage here. Not to mention an almost 6-year old i7-2600K would destroy any FX when both are overclocked, so I'm sorry but I don't see anything to brag about.

Wrong:

Back to the GPU discussion now, bye.

Doesn't look like 100% usage to me. How is talking about CPU bottlenecks not GPU related exactly? It is completely on topic.

Ps: Your Guru3D chart where they tested a very light scene (hence why there's zero CPU scaling) doesn't count.

Except it shows the same scaling as GameGPU does on a 1080... Most GPUs can't outdo the CPU.

This was all in my original post.

 

[dogen1]

No, driver updates are literally hacks to fix issues caused by the developers making calls incorrectly or not optimally.

We shouldn't need driver updates per game release.

Or to replace code that's optimal for one GPU with code that's optimal for another GPU.

Either way, there is a compiler there that does optimize the code, whether you like it or not lol.

 

[Bacon1]

Nvidia released their guide:

Here are some of the major impact features:

Spoiler

(Yes texture resolution does impact performance even on a 1080)

http://www.geforce.com/whats-new/guides/watch-dogs-2-graphics-and-performance-guide

 

[Bacon1]

Computerbase review up as well, they tested HFTS and PCSS as well

70% performance loss with HFTS over Ultra Shadows on 1060

Ultra:

HFTS:

https://www.computerbase.de/2016-11/watch-dogs-2-benchmark/2/

 

[96Firebird]

Ultra shadows are WAY too sharp.

 

[Keysplayr]

No, driver updates are literally hacks to fix issues caused by the developers making calls incorrectly or not optimally.

We shouldn't need driver updates per game release.

I highly doubt that is all driver updates are.

 

[[DHT]Osiris]

Ultra shadows are WAY too sharp.

Yep, looks like a flat 2d render of the 3d tree.

 

[[DHT]Osiris]

I highly doubt that is all driver updates are.

Long, long ago, you got driver updates once a quarter or so, as a few tweaks were made and some bugs sorted out. Not uncommon to find folks with drivers years out of date which functioned perfectly fine. We now see (with NV at least) releasing drivers weekly/every other week, basically dependent on AAA title release tempos. They're basically patching games now, not the cards themselves. In addition, these driver updates can give *wild* swings in performance, like double digit percentages/20+ FPS on some titles.

 

[Arachnotronic]

Computerbase review up as well, they tested HFTS and PCSS as well

70% performance loss with HFTS over Ultra Shadows on 1060

Ultra:

HFTS:

https://www.computerbase.de/2016-11/watch-dogs-2-benchmark/2/

HTFS looks way better.

 

[LTC8K6]

Funny you missed the FX8350 at default being faster than the 3 year younger and more expensive Core i5 6600 14nm Skylake

You mean the i5-6600 4 thread chip that was running 700mhz slower than the FX-8350 8 thread chip and still pretty much tied it?

 

[[DHT]Osiris]

You mean the i5-6600 4 thread chip that was running 700mhz slower than the FX-8350 8 thread chip and still pretty much tied it?

Can we all agree that this is just a peculiarly optimized game, with some parts being more GPU intensive and some parts being more CPU intensive? The benchmarks are all over the place depending on the source, driver versions, settings, etc.