Forza Horizon 4 Benchmarked: Graphics Performance Tested

[CaptainTom]

Please rename the article to graphic cards and 1 cpu benchmark. "5 GHz with 32GB of DDR4-3400 memory" really.. Where's the AMD offering? Using an overclocked chip with 32gb of memory we only, we don't know where the performance floor is. Great a 1060 gets just under 60fps @ 1080p. but who would have a 5gb with 32gb memory and a 1060? That person with an aged box or a stock cpu.. might be mislead. And if it's the argument that the cpu doesn't really matter.. where's that evidence?

as for "GTX 980 looks a little lackluster, getting beaten by pretty much all the mid-range AMD offerings such as the R9 Nano and 390" The GTX980 is 4 years old, the 390 is 3 years old, being on roughly the same tier.

Whats your margin of error? Much of the writing is pointing out how something beat another, yet they're 1 frame faster...

Why introduce more variables? A CPU review will Show what you are looking for.

However, Some conclusions can still be made. Assuming that this game uses all 6 cores (doubtful), 5 ghz is still not enough to bottle neck the 2080 ti. That card gets 160 fps. This means a CPU half that speed will get At LEAST 80 fps. In other words, the only people that MIGHT be cpu bound are those running bulldozer or worse.

Margin of error or not, this was a bad showing for the GTX980.

What's wrong with the 980? It's a 256-bit GDDR5 card with 5TFLOPs. Where should it be?

When will people learn that ROPS only get you so far? Eventually the games get demanding enough that you just need more bandwidth and computational power.

 

[Nightfire]

Please rename the article to graphic cards and 1 cpu benchmark. "5 GHz with 32GB of DDR4-3400 memory" really.. Where's the AMD offering? Using an overclocked chip with 32gb of memory we only, we don't know where the performance floor is. Great a 1060 gets just under 60fps @ 1080p. but who would have a 5gb with 32gb memory and a 1060? That person with an aged box or a stock cpu.. might be mislead. And if it's the argument that the cpu doesn't really matter.. where's that evidence?

as for "GTX 980 looks a little lackluster, getting beaten by pretty much all the mid-range AMD offerings such as the R9 Nano and 390" The GTX980 is 4 years old, the 390 is 3 years old, being on roughly the same tier.

Whats your margin of error? Much of the writing is pointing out how something beat another, yet they're 1 frame faster...

Why introduce more variables? A CPU review will Show what you are looking for.

However, Some conclusions can still be made. Assuming that this game uses all 6 cores (doubtful), 5 ghz is still not enough to bottle neck the 2080 ti. That card gets 160 fps. This means a CPU half that speed will get At LEAST 80 fps. In other words, the only people that MIGHT be cpu bound are those running bulldozer or worse.

Margin of error or not, this was a bad showing for the GTX980.

What's wrong with the 980? It's a 256-bit GDDR5 card with 5TFLOPs. Where should it be?

When will people learn that ROPS only get you so far? Eventually the games get demanding enough that you just need more bandwidth and computational power.

Well, it barely gets 60 fps at 1080p, it loses to the R9 390 (a card that typically targets the GTX 970), and gets crushed by the GTX 1060 6gb.

It looks like Nvidia is giving Maxwell the same treatment as Keplar. Early Retirement.

 

[CaptainTom]

Well, it barely gets 60 fps at 1080p, it loses to the R9 390 (a card that typically targets the GTX 970), and gets crushed by the GTX 1060 6gb.

It looks like Nvidia is giving Maxwell the same treatment as Keplar. Early Retirement.

Kepler fell apart because it had severe weaknesses in its architecture that eventually were not worth programming around, Nvidia didn't nerf anything. Maxwell is much more of a bruit force architecture, but eventually its weakness will become apparent too - for instance HDR causes a massive performance hit because Nvidia cannot rely on its memory compression anymore to make up for its low bandwidth.

The R9 390 is a 512-bit 5.1TFLOP card with 8GB of VRAM. It was always going to pass the 980 eventually - and in fact the 390X will likely pass the 980 Ti soon. Also don't be surprised if in a few years Vega 64 is trading blows with the 1080 Ti in most games.

 

[Nightfire]

Kepler fell apart because it had severe weaknesses in its architecture that eventually were not worth programming around, Nvidia didn't nerf anything. Maxwell is much more of a bruit force architecture, but eventually its weakness will become apparent too - for instance HDR causes a massive performance hit because Nvidia cannot rely on its memory compression anymore to make up for its low bandwidth.

The R9 390 is a 512-bit 5.1TFLOP card with 8GB of VRAM. It was always going to pass the 980 eventually - and in fact the 390X will likely pass the 980 Ti soon. Also don't be surprised if in a few years Vega 64 is trading blows with the 1080 Ti in most games.

It's harder to determine future trends comparing different companies. That has a lot to do with how hard each work on optimizing.

We can see a trend on how each company supports their older hardware. Hawaii falls behind Polaris now, especially with Vulkan. This is due in part to Tesellation bottlenecks. For the most part, older AMD hardware still holds up very well ie Tahiti and Tonga often embarrass Keplar.

Then we compare the 980ti and 1070. These cards were consistently neck and neck until recently. It is like Nvidia wants to convince more of their users to upgrade. There is ABSOLUTELY NO REASON that the 1070 should be 25% faster in Forza and now Assasins Creed. The same thing in regards to the 980 vs the 1060. These pairs have similar specs but for reasons unknown, Maxwell is falling far behind.

 

[CaptainTom]

It's harder to determine future trends comparing different companies. That has a lot to do with how hard each work on optimizing.

We can see a trend on how each company supports their older hardware. Hawaii falls behind Polaris now, especially with Vulkan. This is due in part to Tesellation bottlenecks. For the most part, older AMD hardware still holds up very well ie Tahiti and Tonga often embarrass Keplar.

Then we compare the 980ti and 1070. These cards were consistently neck and neck until recently. It is like Nvidia wants to convince more of their users to upgrade. There is ABSOLUTELY NO REASON that the 1070 should be 25% faster in Forza and now Assasins Creed. The same thing in regards to the 980 vs the 1060. These pairs have similar specs but for reasons unknown, Maxwell is falling far behind.

Pascal actually does have some tweaks in how it handles DX12 commands - it has like half-async support. Maxwell just can't handle modern API's, it's practically a DX11 ASIC.

 

[Nightfire]

Pascal actually does have some tweaks in how it handles DX12 commands - it has like half-async support. Maxwell just can't handle modern API's, it's practically a DX11 ASIC.

Oh come on, Maxwell does dx12 just fine. The only difference is that these were released late 2018.

Furthermore, what is the excuse for AC: Odyssey? Check out Steve's review on Hardware Unboxed. That game uses an older engine, AdvilNext 2.0, which has been around since 2014. But yet again, Maxwell gets a 25% penalty since the game is brand new and Nvidia wants to convince their users to upgrade.

 

[CaptainTom]

Oh come on, Maxwell does dx12 just fine. The only difference is that these were released late 2018.

Furthermore, what is the excuse for AC: Odyssey? Check out Steve's review on Hardware Unboxed. That game uses an older engine, AdvilNext 2.0, which has been around since 2014. But yet again, Maxwell gets a 25% penalty since the game is brand new and Nvidia wants to convince their users to upgrade.

I don't think anyone should be including that in average comparisons of GPU's. I am excited to get ACO, but I will be giving it a month to receive patches and driver updates. Don't be surprised if multiple cards get 40% performance boosts, something is weird.

Heck Ubisoft themselves said they will be patching all 3 versions of the game soon to massively improve performance. FYI- ACIII had this problem around launch as well.

 

[Forebode]

Because it is a GPU benchmark and you use the most powerful CPU so you can avoid CPU boottleneck with the objective to... er... benchmark a GPU?

it states Graphics performance, not Graphic Card performance in the subheading. You bench with a lesser processor to show the optimization instead of talking about possible optimizations.

"to work out what kind of GPU power you’ll need " So it doesn't matter if the game is bound to 1, 2, 4, 4+ cores? Does it not matter if that OC on the processor actually benefits anything or not?

 

[Black Slasher]

it states Graphics performance, not Graphic Card performance in the subheading. You bench with a lesser processor to show the optimization instead of talking about possible optimizations.

"to work out what kind of GPU power you’ll need " So it doesn't matter if the game is bound to 1, 2, 4, 4+ cores? Does it not matter if that OC on the processor actually benefits anything or not?

*Roll eyes* *Sigh*

 

[Nightfire]

it states Graphics performance, not Graphic Card performance in the subheading. You bench with a lesser processor to show the optimization instead of talking about possible optimizations.

"to work out what kind of GPU power you’ll need " So it doesn't matter if the game is bound to 1, 2, 4, 4+ cores? Does it not matter if that OC on the processor actually benefits anything or not?

No one agrees with you here. No respectable reviewer agrees with you. Adding various CPUs into the mix just complicates the data. If you want to see what kind of utilization a game does, check out a CPU review with multiple games tested. If the game is not in there, find one with the same 3ngune and you will have a good idea of what kind of CPU you need.

Bottom line, this is a racing game so it will almost always be GPU bottlenecked unless you do something silly and pair a 2080ti with a celeron looking for max fps at 1080p.

Just let it go.

 

[Nero7]

It would be indeed interesting to see what performance one could get with lets say an i7-2600 and DDR3 RAM. Wouldn't have to be made with all graphic cards just a few of different leagues would suffice.

Anyway, its quite discouraging to see AMD once again triumph over Nvidia. The next cards could really put Nvidia to shame especially since RTX costs so much. Might even force them to bring out a GTX2080.

Also is this using DirectX 12? Or is it dead? Why are we still playing new games with DX11... this is extremely dissapointing and I can only guess that card manufacturers have no interest into DX12 and alternatives because with the FPS boost many people would not need to get new cards.

Lack of PhysX effects and the likes is also very dissapointing. The deformation of snow and sand is nice but really not that great looking. Rather primitive really. As are the objects like rubble from destroyed walls. Car deformation is also absent.

Really sad to see that games are technically stuck at the last gen level and just get higher resolution for both rendering and textures. The rest is stagnating.

 

[NightAntilli]

Poor Fury (X)...

 

[Nightfire]

Poor Fury (X)...

I was a bit surprised here as well. At first, I dismissed frame buffer size as it didn't seem to matter with the rx580.

Then we see the 1060 6gb beat the 3gb by 30% which is more than usual.

More importantly, the Vega56 is a whopping 50% faster than the FuryX. Usually it is closer to a 20% delta. I can see no other explanation to this other than frame buffer size as architecture is similar.

Wolfenstein 2 is the only other game I know where the 8 GB rx580 crushes the FuryX. That game doesn't just ask for more than 4gb, but DEMANDS it. There, we see the Vega56 100% faster than the FuryX.

I suspect lowering the settings to high would make a world of difference for the FuryX. Same goes for the 3gb gtx1060.

 

[videogame57]

"Finally the GTX 780 and 780 Ti for whatever reason locked up when loading the benchmark, I tried many workarounds, all failed."

The GTX 780 and GTX 780 Ti specifically are broken in Nvidia drivers newer than 399.24. They run extremely unstable in all games. I'm not sure why this is and why it only affects these specific cards. Maybe it only affects GK110? But then it would affect the original TITAN too which I haven't heard of having issues.

I almost bought a new GPU because I thought my GTX 780 was failing, but after cleaning the newer drivers and installing 399.24, it works completely fine.

 

[evolucion8]

It seems this game just doesnt like older hardware. Pascal easily beats Maxwell, Polaris outshines Hawaii, and Tonga is much better than Tahiti.

Tonga and Fiji are based on the same architecture and they tend to be Tessellation bound more often than not compared to Polaris. AMD finally managed decent Tessellation performance with Polaris and Vega.

 

[Stark]

No RX 570 8gb ?
testested the 1060 3 and 6gb and RX 580 4 and 8gb but missing the RX 570 8gb for shame
glad to see the 7870 in there, could probably "tweak" some settings to have it run closer to 60 FPS @ 1080p instead of a forced "medium" unless there is not option to adjust things and only use a "flat" setting?

Thank you for the testing, though some of these are not where I expect them to be based on the countless hours of reviews and such I have seen over the years, but kudos if you did in fact run every single one of these cards and not just used a "estimated" given performance grading system ^.^

Not sure where you have seen Vega 56 for less money then a 1080 however, they have almost since day of launch sky rocket in pricing and have yet to be MSRP/SEP from what I have seen?

Even Benchmark King has his limits >.<

 

[Charles Olson]

No RX 570 8gb ?
testested the 1060 3 and 6gb and RX 580 4 and 8gb but missing the RX 570 8gb for shame
glad to see the 7870 in there, could probably "tweak" some settings to have it run closer to 60 FPS @ 1080p instead of a forced "medium" unless there is not option to adjust things and only use a "flat" setting?

Thank you for the testing, though some of these are not where I expect them to be based on the countless hours of reviews and such I have seen over the years, but kudos if you did in fact run every single one of these cards and not just used a "estimated" given performance grading system ^.^

Not sure where you have seen Vega 56 for less money then a 1080 however, they have almost since day of launch sky rocket in pricing and have yet to be MSRP/SEP from what I have seen?

Here in Canada a GTX 1080 Strix is 725 ( plus one free game ) vs PowerColor RED DRAGON Radeon RX Vega 56 is only 579 ( plus there free games ) but this may change for your country !

 

[Brahman05]

Did I miss vram and system ram usage? My 1080 is more than enough but just curious