Older CPU's (no Devil's Canyon)



So, a FX 9590, overclocked to 5Ghz, bumping out 220+ watts of TDP, is outperformed by an I3 4360 at stock speeds...

Quite interesting to note that this game is a port from consoles, including the new generation of consoles powered by 8-core AMD cpu's.

As we can see from the graphs above, the game is clearly well threaded, as the 5960x commands a healthy lead over the 4770k.

The Pentium G3258 also gets annihilated, showing that a dual core without hyper threading is totally useless for new games being released, regardless of what clock speed it's overclocked to (minimum fps of 11 when overclocked to 4.5Ghz)

Here's the CPU core utilization, to confirm the game is well threaded:

I am genuinely surprised that an i3 manages to beat the 8 core AMD CPU's given how well threaded this game is. Usually with heavily threaded tasks the 8350 etc even beats out i5's so an i3 beating them in this I find shocking, that being said rendering is almost exclusively a maximum performance thing i.e all threads are maxing all cores pretty much, this game doesnt do that unless you have 4 threads/cores I see. Need way more tests with i5's since thats what most people run... with games like this and vulkan/DX12 right around the corner, the time for having an i7 might actually make sense for gaming is finally here.

EDIT: There seems to be a huge disparity between these graphs, in the second one the 4330 is beaten pretty solidly by all piledriver 8 cores, yet in the first graph the 4330 is only slightly slower than the 9590? They use different GPU's, second graph with 2 980s and first with 1 970, I guess that shows that the more threads on the AMD CPU's help multi-GPU setups a lot more than something like an i3, makes sense really.

Even though the game is well threaded the draw call execution is still done on one core. Single core performance is still king in draw call heavy games.

So this tells us that if the only game you play is GTA V or any of the other Dave specially "selected" benchmark games, it would be better to get an i3 for slightly better performance over AMD to save a few pennies. Its a fair point dave but you can rule out any sort of OCing going this route though.

Is this initial limitation of the AMD CPUs on this game to do with the "shared" nature of the FX cores do you think? I'm interested to see how the old phenom x6's perform in comparison.

Hang on, these figures don't make any sense. If the CPU was bottle necking, ie: it was the limiting factor in determining performance, you would expect to see it maxing out 90%+ on at least a single core.

That's not the case with any of these CPUs except for the i3, which suggests that something else is restricting performance. What mobo, ram, SSD etc was used for the tests?

AMD CPU's are so bad hope whatever they release next is actually able to beat Intels I3.

Scurburg said:

So this tells us that .. is ... the other Dave specially "selected" benchmark games

Click to expand...

Nothing we are not used to by now.

Concrete said:

Hang on, these figures don't make any sense. If the CPU was bottle necking, ie: it was the limiting factor in determining performance, you would expect to see it maxing out 90%+ on at least a single core.

That's not the case with any of these CPUs except for the i3, which suggests that something else is restricting performance. What mobo, ram, SSD etc was used for the tests?

Click to expand...

I smell something initially which is why an educated calmness is in order till the skew is found. The knee jerk excitement of posting something like this just highlights how amateur some people are in their eagerness to mar AMD.

I do find this somewhat amusing. My Atari ST is better than your Amiga!

Regardless of the results, it's hilarious to see quite possibly the most anticipated PC game in the last few years be classed as cherry picking.

Martini1991 said:

Regardless of the results, it's hilarious to see quite possibly the most anticipated PC game in the last few years be classed as cherry picking.

Click to expand...

Yep, it's quite amusing to see them fretting Now they say there must be something 'fishy' as these results cannot possibly be correct, right?

Welcome to reality, where Intel CPU's are the best for gaming, period.

Not even surprised. There's a reason the FX chips are cheap

Concrete said:

Hang on, these figures don't make any sense. If the CPU was bottle necking, ie: it was the limiting factor in determining performance, you would expect to see it maxing out 90%+ on at least a single core.

That's not the case with any of these CPUs except for the i3, which suggests that something else is restricting performance. What mobo, ram, SSD etc was used for the tests?

Click to expand...

API bottlenecks are often called CPU bottlenecks but won't always be manifest as a core or cores working flatout - the CPU could be feeding the GPU as fast as it is able but not utilising the full extent of its capabilities due to render API inefficiencies.

Also your missing that the i3 is 2 real cores + 2 hyperthreading units - compared to all the other CPUs having atleast 4 real cores - so those 2 real cores have a ton of stuff whacked on them whereas on the other CPUs worker threads, etc. are spread out over the additional cores. (EDIT: As someone alluded to above 1 of those CPU cores is doing the grunt of the draw call execution as well as sharing incidental threading load whereas on the other CPUs its much more dedicated to rendering - which could have interesting implications as the AMDs might be putting up the same or lower performance numbers on paper but potentially could actually be smoother to play on as the cores aren't having to do acrobatics to fit everything in so to speak).

EDIT: For those that malign hyperthreading - that i3 there is showing them so wrong heh.

Oh..that's nice, absolutely no reason to upgrade to a 4790K from my 4690K.

Hyperthreading hype fails once more...

^^ You say that but its certainly working to good effect on that i3 (There is also a fair performance difference between the 2500K and 2600K despite being very close clock speed wise and the only real difference the HT capabilities).

On the higher end CPUs its a mixed story and even when a benefit not always reflected in the raw performance numbers i.e. the i7 might be smoother in intensive situations even though the framerates are largely similar. Generally though the difference between an i5 and i7 don't manifest at 60fps gaming with a single even high end GPU though I do find it useful for overall system responsiveness when playing games where I've got other stuff running as well i.e. teamspeak and/or multiple game clients open. I've found it a little more noticeable when doing 120+fps and multiple GPUs though.

Those results contradict mine. Hrmmm.

Martini1991 said:

Regardless of the results, it's hilarious to see quite possibly the most anticipated PC game in the last few years be classed as cherry picking.

Click to expand...

Actually, I was not saying hes cherry picking, he has in the past though. GTA V is a big deal, which was why I agreed that if this is one of the main games someone wants to play (along with others games that dave likes to highlight) then the i3 IS a valid option for performance in those particular instances ATM, but that would also mean sacrifice things like OCing. I still think it would be rather shortsighted to recommend the i3 over an 8320 though.

Also the game has barely been out for long and its clearly showing a preference for 4 intel cores, and as I was asking, is this an issue with the way the amd cores share resources OR is it an optimization issue, or both. Does limiting the fx's to 1 core per shared resource help or have any affect on the results?

I've been trying to explain this for ages... some people have forever been saying that upcoming games will run brilliantly on AMD due to the nature of consoles being 8 core and games as a result being heavily multithreaded, the fact is though the 8 core CPU in the consoles is so pathetically weak that even an i3 has no trouble executing those same threads/workloads in less time. In fact, it looks as though the i3 only starts to become a bottleneck with 980SLI and that could well be largely down to API.

DX12 is probably going to mean that even a basic quad core (eg. 4690K) is overkill for gaming so I think buying an AMD 8 core with future gaming in mind is misguided. Faster cores will ALWAYS produce higher frame-rates in non-bottlenecked situations.

Look like decent performance increases from overclocking a 4770k anyway.

All I get from those graphs is that I'll be able to play GTA V on my 8320 with decent fps.

Sweet.

GTA 5 isn't a port from consoles. It's this thinking that's the whole problem with people thinking that games are ported. They get ideas in their head about performance, issues, and bugs then can't help but blame it on consoles.

Has anyone found some CPU benchmarks at more sensible settings with weaker/older CPUs? Before buying GTAV I want to make sure the game does'nt have to be turned down too much to get decent performance with my Phenom II X4 955.