why are amd processors so much slower than intel in games

[etrin]

I know there is a speed difference but is that it or is it the processor, motherboard or both that cause the big difference ?

 

[dave_the_nerd]

They're not THAT much slower - in most cases, at a given price point, they perform about the same.

Intel pulls away a little at the high end, in games that prefer higher single-threaded performance to better multithreaded performance. For games with better multithreading, it's actually the opposite.

RAM speed matters more now than it used to, since RAM speed is the base clock used for the interconnect between the two halves of a Ryzen CPU. Some people are probably buying slower RAM than they should and getting hamstrung a bit as a result.

The ongoing GPU shortage is a bigger problem for gamers.

 

[bfun_x1]

Not that much slower and not always slower. Ryzen also tends to have better frame time variance which is arguably more perceivable than the slightly lower FPS.

 

[ao_ika_red]

Which processors are we talking about? If it's FX-8350 vs i7 3770k, yes indeed. The performance gap was as wide as grand canyon valley. But, ever since Ryzen came, it's much narrower. Here's the example:
https://www.anandtech.com/bench/product/1950?vs=1543

 

[IllogicalGlory]

There are a few reasons.

1. Clock speed - OC to OC, Intel chips have about a 20-25% higher clock speed compared to Ryzen, which has a large effect in some games.
2. IPC - Even in applications that don't care about CCX latency (Cinebench, et. al.), Intel gets about 10% more performance for each clock cycle. So more clock cycles per second and more instructions for each clock cycle on Intel.
3. CCX communication latency - the CCX design makes communication across them slower than mainstream Intel chips, which use the ring bus. Skylake-X, with its mesh topology is similarly disadvantaged. Some games don't care much about this, but others are fairly heftily affected.
4. Memory latency - Intel CPUs take about 40-60ns to access memory, while on Ryzen, it's more like 80-100ns.

These factors will have varying effects depending on the game.

To answer your question about processor vs motherboard, it's all on the processor. Motherboards don't affect performance directly, but some will allow higher overclocks and higher memory speeds. However, neither X370 or Z170/270/370 motherboards have any perceptible advantage over the other in this regard.

 

[Snowleopard3000]

I have not had a chance to try Ryzen or Threadripper, but my experience with AMD and Intel can be summed up like this.....

Intel is like a Dodge Viper. They over kill on the horsepower, but deliver it horribly to the wheels. AMD is like a finely tuned Porsche or Ferrari. Then come the real world tests because how many people only drive their cars on perfectly paved race tracks vs. normal pot hole filled, cracked pavement?

In my experience, there is nothing wrong with AMD Processors, but when the crap hits the floor and I need stuff to get done, AMD processors never have made it, so I stopped using them,

Don't take these numbers as exact figures, but as an example, I need 1.5 AMD processors for every 1 Intel processor to get the same amount of work done.

Then AMD processors are usually cheaper than Intel Processors, but I am getting less with the AMD processor in terms of raw i/o or raw power.

Again this is before Ryzen and Threadripper came out. So now I need to re-evaluate my next pc builds based on the new AMD processors available.

 

[whm1974]

Other then running benchmarks, could we really tell the difference between Ryzen and Coffee Lake?

 

[prtskg]

Other then running benchmarks, could we really tell the difference between Ryzen and Coffee Lake?

Yes. One of them feels smoother in games because of more number of good cores for the same price OR so some people say.

 

[bbhaag]

Other then running benchmarks, could we really tell the difference between Ryzen and Coffee Lake?

I know I can't. My 40yo eyes are well past the point of being able to tell the difference between 80fps and 83fps in a game or the ability to differentiate between 30 seconds and 30.12456 seconds when unzipping a file.haha

 

[ao_ika_red]

I know I can't. My 40yo eyes are well past the point of being able to tell the difference between 80fps and 83fps in a game or the ability to differentiate between 30 seconds and 30.12456 seconds when unzipping a file.haha

+1 Without FPS count overlay or FreeSync /G-sync, I can't tell the difference past 60 fps mark.

 

[DigDog]

frame time variance

???
https://www.youtube.com/watch?v=yV7wm1FbhS4

 

[arandomguy]

Other then running benchmarks, could we really tell the difference between Ryzen and Coffee Lake?

Depends on the usage scenario, SKUs being compared, and system configuration.

I'm still of the opinion that a lot of CPU based gaming benchmarks are misleading as they take too small a time slice of any given game. So if anything benchmarks might be under representing some of the differences.

Also I feel reviewers should explore more multitasking while gaming. I know the reasoning behind running as lean as possible for bench marking but at least to me that isn't how modern users actually use their PCs. Or maybe I and the circle of people I've seen is the actual minority but I certainly leave quite a lot of background activity even when gaming. This isn't even referring to something such as encoding via the CPU.

 

[IRobot23]

1. 6% ipc (not so important)
2. 20%+ core clock (not so important)
3. 2x memory latency (very important)

Skylake X shows that memory latency is important (gaming) and with same clock and same memory latency they actually lagg behind AMD Ryzen. Zen is very capable core.

Benching faster memory:
DF, HU, GN etc. did worst job ever. AMD profiles and x299 profiles for dram are basically "broken".

Anyway, if you get your ryzen near 65ns(aida64 mem&cache), I am sure that it will be a beast.

 

[mikk]

Which processors are we talking about? If it's FX-8350 vs i7 3770k, yes indeed. The performance gap was as wide as grand canyon valley. But, ever since Ryzen came, it's much narrower. Here's the example:
https://www.anandtech.com/bench/product/1950?vs=1543

But only because it's GPU limited. If it's not bottlenecked by the GPU Intel pulls away. If you want the best gaming performance you have to buy an i7-8700k with high OC DDR4.

 

[ao_ika_red]

But only because it's GPU limited. If it's not bottlenecked by the GPU Intel pulls away. If you want the best gaming performance you have to buy an i7-8700k with high OC DDR4.

Truth to be told, I never say that Ryzen is equal with Intel's current offerings (that's CFL family). Ryzen supposed to fight with Skylake, but it came too late. By then, Intel had fine-tuned Skylake in form of Kabylake and Coffeelake. I still even doubt if Ryzen's next iteration will equalise its performance with CFL but because the performance gap much closer than in Bulldozer era, it becomes a more sensible option to customers.

 

[IRobot23]

But only because it's GPU limited. If it's not bottlenecked by the GPU Intel pulls away. If you want the best gaming performance you have to buy an i7-8700k with high OC DDR4.

I switched to R7 1700 from i7 3770. And r5 1500x destroys it.

 

[DrMrLordX]

when the crap hits the floor and I need stuff to get done, AMD processors never have made it

Obviously, you weren't around for the Thunderbird, various Athlon XPs, and of course K8. AMD was wiping the floor with Intel in that timespan except for when the Northbridge-C chips briefly retook the performance crown. Once K8 caught on, it was over with until that fateful day in 2006.

AMD held the desktop 1P performance crown for most of the time period between 2001 - 2006. You sure AMD processors never made it?

 

[whm1974]

Obviously, you weren't around for the Thunderbird, various Athlon XPs, and of course K8. AMD was wiping the floor with Intel in that timespan except for when the Northbridge-C chips briefly retook the performance crown. Once K8 caught on, it was over with until that fateful day in 2006.

AMD held the desktop 1P performance crown for most of the time period between 2001 - 2006. You sure AMD processors never made it?

I agree, I had a Duron, Thunderbird, and AthlonXP between middle 2000 to 2006. Better performance then Intel at the time and much cheaper.

 

[scannall]

The title seems like a troll to me. Define "So much". And in what circumstance? They are somewhat slower in 1080p, super fast monitors with low settings.. Not a lot, but some. If your monitor is 60hz, there is no difference. Or at 1440 and above, no difference. I'd add, we don't know what the end result of the meltdown patches and microcode updates are going to do yet.

 

[bfun_x1]

???
https://www.youtube.com/watch?v=yV7wm1FbhS4

https://www.youtube.com/watch?v=ybF7r4rogHc&feature=youtu.be

 

[MajinCry]

It's due to draw call performance. There's a significant penalty when the driver thread is on a separate CCX than the game threads.

https://forums.anandtech.com/threads/part-2-measuring-cpu-draw-call-performance.2499609/

This benchmark is a raw, draw call monster. It shows how efficient each architecture is at processing draw calls, removing the caveats that come with games (i.e, how well coded it is, and how well threaded).

Even with very fast RAM, the inter-CCX performance penalty is still pretty big.

 

[DrMrLordX]

It's due to draw call performance. There's a significant penalty when the driver thread is on a separate CCX than the game threads.

BD had no CCX penalty, yet they were slower than Intel chips in games at the time . . .

 

[PeterScott]

They're not THAT much slower - in most cases, at a given price point, they perform about the same.

Intel pulls away a little at the high end, in games that prefer higher single-threaded performance to better multithreaded performance. For games with better multithreading, it's actually the opposite.

I agree that it isn't really that big of a deal in the Ryzen generation (It was in the Dozer generation).

But don't correct one inaccuracy by stating another. Intel works just fine with multi-threaded games.

What games run faster on Ryzen than Coffee Lake?

 

[MajinCry]

BD had no CCX penalty, yet they were slower than Intel chips in games at the time . . .

I assumed OP was talking about Ryzen. You can see the inter-CCX penalty lowering the draw call perf in the benchmark.

For Excavator and older AMD architectures, they're just really bad at draw calls. Couple that with lower raw processing power than their competition, and you get poor game perf.

 

[NostaSeronx]

BD had no CCX penalty, yet they were slower than Intel chips in games at the time . . .

For Excavator and older AMD architectures, they're just really bad at draw calls. Couple that with lower raw processing power than their competition, and you get poor game perf.

Bulldozer has slow interconnect speed, no register renaming, and used execution cores based on pre-2006 designs. Modified POWER4 & Alpha 21264 because Charles R. Moore, rather than the Mitch Alsup derived K9 core with CMT.