The 8 core showdown and analysis thread.

[Wraith]

As it stands we all know bang for buck you can't go wrong with AMD and you'd be foolish not to consider one, if you have the wedge though then Intel is always the way to go. Nice write up by the way V-T I at least appreciate the efforts and the point your trying to make, but you can't rightly compare a 8 core 16 thread server chip to a generic 8 core desktop die especially given that the FX was OC'd.

Intel always had the edge up until recently given their dirty marketing tactics, buying out suppliers and manufacturers.. the future is where it's at now. I'm hoping that rumors are true of Samsung wanting to aquire AMD, we'll see some real battles in the CPU future if this happens.

EDIT: Before I get a rage reply, I DO NOT condone FLAMING of any sort, but you must have known that you'd get some raised eye brows and replys.

 

[MadShadow]

As it stands we all know bang for buck you can't go wrong with AMD and you'd be foolish not to consider one, if you have the wedge though then Intel is always the way to go. Nice write up by the way V-T I at least appreciate the efforts and the point your trying to make, but you can't rightly compare a 8 core 16 thread server chip to a generic 8 core desktop die especially given that the FX was OC'd.

Intel always had the edge up until recently given their dirty marketing tactics, buying out suppliers and manufacturers.. the future is where it's at now. I'm hoping that rumors are true of Samsung wanting to aquire AMD, we'll see some real battles in the CPU future if this happens.

EDIT: Before I get a rage reply, I DO NOT condone FLAMING of any sort, but you must have known that you'd get some raised eye brows and replys.

+1 on this. V-T, you clearly put some good work into this, but you really cant compare the two chips. Well you could... But you see what I'm saying

 

[SPS]

I'm actually quite surprised how well the low-clocked Intel performed compared to an AMD chip running at 4.9GHz.

Very few games (and a lot of apps) will scale to 16 threads at any one time during normal use/mid gameplay, this means that this test boils down to single threaded performance. The AMD chip therefore takes the crown most likely down to its clock speed. An interesting test would be to repeat the test while decrementing the AMD clock speed to see at what speed the Intel takes the win.

Games usually have a game thread and a render thread, some engines also make use of render worker threads to perform things like resource preparation, animation processing using DX11's deferred contexts/command lists, the final rendering submission can still only be done on one thread using the immediate context though. (Other threads are used for non-rendering tasks of course). We can see that the difference in GPU scores for Intel on 3DMark11 and Firestrike was ~700 and ~30 which means that there must have been some threading improvements.

The physics score is lower for Firestrike (than the AMD) which is an odd one because it simulates 32 separate bullet worlds, one thread per CPU core. I can only assume it used the physical core count only..?

Anyway, there's some interesting data there, thanks for posting.

 

[Wraith]

Games usually have a game thread and a render thread, some engines also make use of render worker threads to perform things like resource preparation, animation processing using DX11's deferred contexts/command lists, the final rendering submission can still only be done on one thread using the immediate context though. (Other threads are used for non-rendering tasks of course).

Hey V-T have you tried S.T.A.L.K.E.R call of pripyat, for benching? I'd actually like to see how the single core gaming performance goes between these 2, just for raw power as it's one of the most CPU intensive games for single - dual core systems. And if you want a real multi-thread face off then it has to be Arma III. Now that would give some real eye opening results.

 

[barnsley]

And if you want a real multi-thread face off then it has to be Arma III. Now that would give some real eye opening results.

Arma III is not a good example of multi-threading in the slightest imo. It does not actually use the four cores that you will see activity on properly. Fast cores and very fast storage is more important. I'll try and produce a graph thing that you can get within game to show what part of the game is using cpu core wise.
-edit- I can actually show without needing to change rigs within game as I can force it to use only a certain number of CPU cores.


If you ask me, Battlefield does it much better.

 

[Wraith]

Arma III is not a good example of multi-threading in the slightest imo. It does not actually use the four cores that you will see activity on properly. Fast cores and very fast storage is more important. I'll try and produce a graph thing that you can get within game to show what part of the game is using cpu core wise.
-edit- I can actually show without needing to change rigs within game as I can force it to use only a certain number of CPU cores.


If you ask me, Battlefield does it much better.

Strange that, because on the FX-4100 it utilised all 4 cores smoothly, still I'd be interested to see how the vashera handles it.

 

[barnsley]

Strange that, because on the FX-4100 it utilised all 4 cores smoothly, still I'd be interested to see how the vashera handles it.

Quick test i did using a 40 AI custom map I made quickly shows that I was getting a max of 44fps and an average of 40fps in it (graphical settings were on ultra but the gpu was not maxing out) using 4 cpus with hyper threading enabled. dual core, no hyper threading gave me a max of 42 and an average of 38. I suspect the fps gained previously was down to hyperthreading being forced enabled. I'd do other settings in a dedicated thread some time but the hijacking stops here

 

[Dicehunter]

Cool tests dude.

It's kind of bad that a cheap AMD chip, Albeit a good chip, Can keep up with an Intel chip that costs literally 10 times more in all the tests you did.

 

[NeverBackDown]

I don't find any of this data or results prove anything. Not impressed tbh.

It really just confirms that a sever chip running at 2ghz is worse in a few areas than a desktop class chip running at nearly 5ghz.. It also confirms what we already know.. that price/performance the 8320 is the best cpu(at its price range). Other than that i can't really agree with this thesis that appearently most of it was a joke and meant to not be taken seriously.. which then invalidates and dismisses everything else..

Not trying to start anything here but honestly a more even comparison is an i7 4790k. Or even the ivy i7 if you desire. Clock the FX to nearly 5ghz and then compare them with the intel at a set stock clock speed.. then an OC'd result. That is a more even comparison and if you don't think so, get the i5. Stock and OC etc. AMD are good chips, intel just make better ones. Only time will tell with W10/DX12 and Vulkan if AMD can come back and compete again when it allows for better multi threading and etc.

 

[tinytomlogan]

Ive played with dual intel xeons vs a single of their desktop counterpart and games really like the clockspeed a great deal more than youd think.

A reasonable overclock on a single CPU absolutely desimates anything xeons can chuck at them. Xeons are all server and heavily threaded workstation apps like rendering photos/video and game dev.

Ive got two 18 Core Xeons here and Intel dont even want them put through games tests because thats "not the target market"

Did make some interesting reading when comparing it to AMD though when for the mainstream user there is absolutely no difference.

This whole article does just add weight to my "the EVGA SRX is pointless as a 'gaming board' " argument

 

[Dicehunter]

Ive played with dual intel xeons vs a single of their desktop counterpart and games really like the clockspeed a great deal more than youd think.

A reasonable overclock on a single CPU absolutely desimates anything xeons can chuck at them. Xeons are all server and heavily threaded workstation apps like rendering photos/video and game dev.

Ive got two 18 Core Xeons here and Intel dont even want them put through games tests because thats "not the target market"

Did make some interesting reading when comparing it to AMD though when for the mainstream user there is absolutely no difference.

This whole article does just add weight to my "the EVGA SRX is pointless as a 'gaming board' " argument

Just imagine if we could have mobos with proper support for either dual 4790K's or 5960X's :drool:

 

[AlienALX]

+1 on this. V-T, you clearly put some good work into this, but you really cant compare the two chips. Well you could... But you see what I'm saying

Well you can because I did but yeah I did mention it was done 'just because I could'.

Don't worry, next one will be equally as bonkers but interesting. Pentium anni with a Titan black vs 3970x with a Titan black.

I prefer to write about oddball things it's much more fun than writing reviews

I'm actually quite surprised how well the low-clocked Intel performed compared to an AMD chip running at 4.9GHz.

Yup I did point that out. It just makes the padlock stink all the more.

Hey V-T have you tried S.T.A.L.K.E.R call of pripyat, for benching? I'd actually like to see how the single core gaming performance goes between these 2, just for raw power as it's one of the most CPU intensive games for single - dual core systems. And if you want a real multi-thread face off then it has to be Arma III. Now that would give some real eye opening results.

I own all of them I'm pretty sure. Pripyat is the one with the storm clouds and red storms yes?

Cool tests dude.

It's kind of bad that a cheap AMD chip, Albeit a good chip, Can keep up with an Intel chip that costs literally 10 times more in all the tests you did.

Yeah that was pretty much the point I was making. It just shows how horrific the value is on the more expensive Intels.

 

[Wraith]

I own all of them I'm pretty sure. Pripyat is the one with the storm clouds and red storms yes?

It sure is Under the G3258 at 4.0Ghz I was getting solid single core performance, but with the i5 4440 the performance drop is so noticeable even with a 700Mhz difference.. could well be due to the turbo cache as it's not overclockable, but that said the FX-4100 even though it was a quad did a much nicer job at 4.0Ghz in single core tasks. With that in mind I'd 100% in this case go with AMD.

 

[AlienALX]

It sure is Under the G3258 at 4.0Ghz I was getting solid single core performance, but with the i5 4440 the performance drop is so noticeable even with a 700Mhz difference.. could well be due to the turbo cache as it's not overclockable, but that said the FX-4100 even though it was a quad did a much nicer job at 4.0Ghz in single core tasks. With that in mind I'd 100% in this case go with AMD.

Yeah I actually liked that game but a hard drive died as I was getting right into it and sadly I never returned. It's total frickin' hard core though man, no place for the weak. Might be a great game for the new mini rig !

As much as people say bad stuff about AMD most of it is complete nonsense. I deliberately chose the FX 8320 over the at the time 3570k because put simply it suited my needs better. I was running 670s in SLI and soon after a 7990 and in Crysis 3 my frame rates were terrible. At that time I had a Xeon similar to the I5 2400 and I was seeing mins of 22 FPS and worse. Switched out to the AMD and the game flew along, 39 FPS min.

Even the I5 couldn't match the AMD.

There are caveats of course. Power consumption is high (about the same as a I7 950, no one cared then !!!!!) and whilst not particularly hot (72 is the dead zone and will lock the rig) they do create heat fast. As such you need tip top cooling (h100 or go home) and you need a bloody good board and they're not cheap. They have dipped in price ever so slightly (so around £110 now not the £150 I paid) but yeah, you do need a bloody good board with bloody good VRMs and cooling.

I've seen many people make the mistake of buying a 4+1 board and then finding it throttles like crap (1.7ghz). I would laugh but hey, happened to me. I assumed because the Asrock board I bought was £100 and a 990 that it would be fine. Thing was a joke, 4.2ghz was the absolute limit and my H60 fell apart.

But get them threading? they're amazing. I did mention that I see no difference in games (other than GPU horsepower allowing me better MSAA on the Titans) compared to my 3970x and I absolutely mean that and will stand by it. I don't run frame counters in games I simply use my brains and my intuition. If something ain't right believe me, I will spot it.

But that FX is an amazing CPU. It's better than ever now because AMD are chasing 5ghz for the Centurions, so you usually find even the 8320E likes 5ghz as long as you can supply the voltage.

I gave mine to my lady. It's my all AMD rig, 7990 and 8320 and I've never heard her complain.

Tell you what though.. My Alienware Area 51 ALX (the vent top one) is still around. I threw a Dell workstation board and a ES Westmere clocked to 2ghz in that and even that puts up a damn good show in games.

People do not realise that CPUs are just not important any more. Games are being coded to work with a 8 core crap clocked AMD and any other CPU will need optimising. And why bother wasting time and money when you can simply re-use that architectural code when implementing PC code.

It's kind of like a cheat's short cut. In the Dying Light bench every CPU pretty much sat in harmony. You really, truly, don't need a good CPU for these new console slopped games. I won't say ports because logically they are not, but yeah, any old CPU will do you going into the future the race is definitely over and done with.

All about your GPU now. Once these new APIs turn up that will lay testament. The CPU is a dying breed. And about fricking time too !