Digital Foundry: all questioned AAA developers recommend AMD CPU's for gaming PC's

[frozentundra123456]

Some hide this behind the scenes other do not:

http://www.intel.com/content/www/us/en/gaming/overclocking/games-optimized-for-intel.html

Yes skyrim is one of them. Someone doubt it?

Again, what is your point? You seem to imply that it will be something great if/when games are optimized for AMD processors, but that it is somehow cheating when it is done for intel processors.

If software is suddenly optimized for HSA and AMD processors, are you then going to go on the forums and say AMD is cheating because the software is optimized for their architecture?
Didnt think so.

 

[Exophase]

I asked you how the games are being optimized for Intel CPUs, not for Intel's advertising that could mean anything. If that isn't clear, I was wondering if you had any kind of technical description of what that could mean. And that I wasn't interested in claims that they're compiled with ICC unless you can show hard evidence of that.

 

[galego]

Snip

Intel can do that. Who said the contrary? But maintain in mind that consoles will continue to use AMD hardware during more years. Of course, if Intel manages to obtain a competitive CPU then future recommendations will include them. Today all game developers recommended AMD.

Overgeneralize much?? The article only says "a number". They dont even say what that number is, much less it being "all" developers. You correctly quoted the article in the earlier part of your post, but managed to extended it to fit your own agenda in this final sentence.

In the part of my message that you sniped I wrote:

it is understandable why all the game developers who were asked by eurogamer selected the AMD FX-8350 chip as better gaming chip.

I don't need to repeat the bold part in each subsequent phrase or paragraph. It was stated at the first. No confusion is possible unless you deliberately snip that part.

 

[frozentundra123456]

Overgeneralize much?? The article only says "a number". They dont even say what that number is, much less it being "all" developers. You correctly quoted the article in the earlier part of your post, but managed to extended it to fit your own agenda in this final sentence.

In the part of my message that you sniped I wrote:



I don't need to repeat the bold part in each subsequent phrase or paragraph. It was stated at the first. No confusion is possible unless you deliberately snip that part.[/QUOTE]

If you are going to highlight a portion of a sentence by bolding it, you should at least make the sentence correct.

 

[frozentundra123456]

To compare CPUs look at lower resolutions. Check the Crysis 3 benchmark on my post above. The FX is only beaten by a 12-thread Intel chip.

Fine, then look at the low res benchmark in post 12. It shows a stock 3570k decisively ahead of an FX8350 overclocked to 4.7ghz.

 

[galego]

I asked you how the games are being optimized for Intel CPUs, not for Intel's advertising that could mean anything. If that isn't clear, I was wondering if you had any kind of technical description of what that could mean. And that I wasn't interested in claims that they're compiled with ICC unless you can show hard evidence of that.

I provided you a list of Windows games that the developers affirm are optimized for Intel. This list being stored in Intel website does not change anything.

Those are closed source games, therefore unless you are member of the developers team and you have the highest credentials you cannot access to the code and see on what consists the optimization.

 

[Sleepingforest]

Hey, look at AMD's blog! It says thing like "Optimized for AMD: BioShock Infinite launches" and "Crysis 3 runs best on AMD Radeon, and it’s out today!"

They must be cheating the market to stiff Nvidia! And they aren't stopping there! They've optimized FirePro for 3Design, optimized more than 600 hardware and software products since 1999, and even collaborating with companies to optimize software just for their APUs!

How can we stop this? Oh wait. We don't need to. Both companies look for developers to partner up and optimize their software.

 

[AnandThenMan]

There is a vast difference between optimizing, and doing a hardware query and downgrading the compiler code path.

 

[Sleepingforest]

There is a vast difference between optimizing, and doing a hardware query and downgrading the compiler code path.

Where does the link Galego provided show that those things?

 

[AnandThenMan]

Where does the link Galego provided show that those things?

Doesn't change anything, Intel has in the past (and possibly to this day) purposely crippled their compiler code path for no other reason than when it detects a non genuine Intel processor, even if the competing CPU has the appropriate hardware capability. IMO Intel is free to do this, but they must disclose that they do.

 

[galego]

Fine, then look at the low res benchmark in post 12. It shows a stock 3570k decisively ahead of an FX8350 overclocked to 4.7ghz.

The benchmark I gave is for "Welcome to the jungle". The pclab that you give is for "the root of all evil". The guys at pclab obtain the same conclusion for "Welcome to the jungle": The FX ahead of both the i7 and the i5.

It seems that there is some problem with the floating-point computations on the "The Root". It says that the game is in no way optimized for the FX chip and that the benefit over the intel is purely due to the heavy multi-threading.

The review makes an interesting question. It asks if the AMD FX will be more efficient than the quad core intel in the coming games, and then adds that Crysis 3 provide a foretaste of what developers can show in the coming games. Therefore instead disproving what Eurogamer says PCLab review also seems to support the point of AMD CPU shining for future gaming.

 

[Sleepingforest]

Doesn't change anything, Intel has in the past (and possibly to this day) purposely crippled their compiler code path for no other reason than when it detects a non genuine Intel processor, even if the competing CPU has the appropriate hardware capability. IMO Intel is free to do this, but they must disclose that they do.

The link I showed was the reason for my post. I accept that Intel cheated, and should disclose this more publicly, and really shouldn't have done it to begin with. But the particular link Galego provided was meant to prove more of Intel's corporate evil, and I felt like providing a counter-point for that.

 

[AnandThenMan]

But the particular link Galego provided was meant to prove more of Intel's corporate evil, and I felt like providing a counter-point for that.

Fair enough, I appreciate that.

I've always thought that game dev relations and the like are a slippery slope, but it is not going away anytime soon, quite the opposite. It's a double edged sword, Intel helps software devs get the most out of the hardware, but the temptation to handicap your competitor is too great. Intel has done it for at least a decade, Nvidia does it, AMD I am not so sure but if they don't, from a business perspective they should.

I don't like seeing it, but it is what it is.

 

[galego]

There is also a vast difference between optimizing and publicizing it to customers, and maintaining it all in secret, giving customers the impression that some chips are infinitely better than others.

 

[galego]

Doesn't change anything, Intel has in the past (and possibly to this day) purposely crippled their compiler code path for no other reason than when it detects a non genuine Intel processor, even if the competing CPU has the appropriate hardware capability. IMO Intel is free to do this, but they must disclose that they do.

Intel continues doing it, except that the FTC obliged them to add a public disclaimer about how the compiler 'works' with non-intel chips.

 

[Sleepingforest]

I admit that Intel was using unfair practices to get an edge. They evidently still are.

AMD and Intel, on the other hand, both try to get in with developers to optimize games, which is overall very good for the consumer because we get games that are better made for pretty much any platform we want. They promote their works with developers because it shows what apps will get the best performance. This is one of the great impacts of competition that I fear will disappear if AMD does.

 

[galego]

But the particular link Galego provided was meant to prove more of Intel's corporate evil, and I felt like providing a counter-point for that.

It is very difficult to accept that a link to a public partnership program was proving corporate evil. Last time I checked partnership was not illegal...

The link was cited by other reasons.

 

[Sleepingforest]

Oh, sorry, I misunderstood you then. What was it for?

EDIT: Okay, I understand now. Sorry for that attack, but it should be said that this sort of optimization happens on both sides.

 

[Ayah]

But I never said that an FX-8350 would be better than an IB i7. My point, and perhaps I missed the point of your initial post, was that if something is optimized for an 8-core Jaguar, it will run fine on an FX-8350 because each thread of an FX-8350 is more powerful than each thread of an 8-core Jaguar.




Competing CPUs with different architectural characteristics are not new (you mention many that are quite old now). How is this any different? It is an explanation for the differences in performance we see, but actual performance is still all we care about, or at least all I care about.

Furthermore (also see below), each Jaguar core has the same amount of FP resources as a Piledriver core (though the Piledriver core has greater peak FP resources). Additionally, Jaguar is 2-wide, while Piledriver is also (on average ) 2-wide, peak 4-wide rolleyes.


Now, given all this, the big question that I haven't attempted to address (and what people seem to be arguing around) is whether a 4-thread IB i5 will be superior to an 8-thread Piledriver. The answer to that is... who knows :biggrin:. My personal guess is "not any time soon", but I imagine most games will have 4~6 (at the extreme high end) heavy cores, so singlethread performance will probably still win the day. If games can really fully use 8 threads, we already know the results FX-8350 vs i5 when all 8 cores are fully utilized.



Yes, his FX (assuming it is one of the 8000 series) can.


Max theoretical FP throughput is much lower than 8 SB or IB cores, though.

The flex fpu thing is only for applications which don't support AVX. Both AMD and Intel have 256-vit wide FPUs in current gen.

 

[SlowSpyder]

PS4 is going to use eight Jaguar cores, right? Jaguar isn't meant to be a workhorse, it is meant to sip power and be efficient. So for developers to make next gen games, I think they'll have no choice but to create better multithread game engines... they won't be able to use one or two cores and call it a day. So I can see how i7 and FX8xxx may age better than other CPU's, but I think that's still a while off.

 

[inf64]

What most people fail to count in is the fact that neither AMD nor intel will stand still. Intel has Haswell launching soon (not a major speed bump though) and AMD has SR based APUs coming in late 2013 or early 2014. FX8350 will probably stay the fastest AMD chip until mid 2014 (MT performance being deciding factor for that) but 6T SR should be better gaming chip as it should offer more IPC and comparable clocks. On intel side we will have Haswell and IB-E so they will definitely have covered the 8T+ game engines (from PS4/Xbox) when/if these launch on PC market.

 

[galego]

PS4 is going to use eight Jaguar cores, right? Jaguar isn't meant to be a workhorse, it is meant to sip power and be efficient. So for developers to make next gen games, I think they'll have no choice but to create better multithread game engines... they won't be able to use one or two cores and call it a day. So I can see how i7 and FX8xxx may age better than other CPU's, but I think that's still a while off.

The design of the PS4 was developers driven. Sony asked to game developers if they would prefer a 4-core or a 8-core. The 8-core design on the PS4 was developers choice.

 

[Exophase]

The flex fpu thing is only for applications which don't support AVX. Both AMD and Intel have 256-vit wide FPUs in current gen.

Ivy Bridge has an FPU with a 256-bit FMUL pipe and 256-bit FADD pipe. Most of the other functional units are 128-bit.

Piledriver has an FPU with two 128-bit FMA pipes. We could talk about other pipes (integer, shuffle, load/store, etc) but they're also 128-bit. AVX is split into two macro-ops, whether or not they're forced to execute at the same cycle (or "fused" as AMD puts it) isn't really relevant.

Jaguar has a 128-bit FMUL and 128-bit FADD pipe, this is what podspi means when he says it has the same number and width of execution resources as PD.

I think this is very relevant when talking about stuff coming from consoles because it'll have a lot of heavy lifting on the FPU/SIMD units, but won't have a lot of specialization for FMA and XOP, outside of what recompilation gives you. It also won't have a good reason to use AVX256 even though it can, but that might not stop people..

 

[Enigmoid]

The design of the PS4 was developers driven. Sony asked to game developers if they would prefer a 4-core or a 8-core. The 8-core design on the PS4 was developers choice.

Yes, When you ask the devs, "Hey do you want more cpu power?" They are really going to say, "Nope ".

4 core jaguar at 1.6 ghz would be a really bad bottleneck.

 

[galego]

Yes, When you ask the devs, "Hey do you want more cpu power?" They are really going to say, "Nope ".

4 core jaguar at 1.6 ghz would be a really bad bottleneck.

Who said you that is was 8 core jaguar at 1.6 ghz vs 4 core jaguar at 1.6 ghz?