Best Gaming CPUs For The Money: January 2012 (Archive)

Page 60 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.


That they should re-test with DDR3-1866CL8, DDR3-2133CL9, and DDR3-2400CL10 across the board. What I have seen is that these Athlons (and the A series APUs in general), when paired with fast low latency ram (and adequate nb clocks), more than make up for not having L3 cache.

As far as Phenom II goes, it can still brute force through many floating point scenarios where modern AVX instruction paths are not present to allow the BD/PD integer cores to calculate the floating points, and Skyrim is an extreme case of DX9 not scaling properly on modern multi-core platforms. Some community members have modded Skyrim's coding to fix a lot of the x87 trash.

 
Interesting. It's too bad then that you don't see the 750/760k properly reviewed anywhere.

What would the gaming performance increase be for these titles in moving from 1866 to 2400?

One of my personal PCs in the house is pretty much dedicated to a heavily modded Skyrim install with the Phenom 2 at 4Ghz and R9 280x GPU. I was doing fine with that until I upgraded to a 1440p monitor. I'm now needing a platform upgrade to an i5 and 290x. Wish I had a good upgrade path with FX, but it's just not there.
 

Yep - had read that, but that is for integrated graphics. I was looking for something showing impact on discreet graphics.

Looks like damric has done this for 760k for discrete using an HD 7850GPU here:
http://www.techpowerup.com/forums/threads/comparisons-of-cpu-gpu-ram-overclocks-and-comparing-bottlenecks-fm2-cpu.196940/

That's almost an 18% increase in FPS just by taking the 760k from 1600Mhz RAM to 2400MHz (last 2 charts in that post). Impressive!
 
Wow... serious AMD Bias in this article, I mean... an Athlon II x4? Really???

Tom's if you're going to mention those POS CPU's, then at least throw the Core2Duo and Quad in there. An e8600 completely smashes the Athlon II x4. ANY CORE 2 QUAD will put any Athlon II CPU to shame. Do you guys at Tom's hate the Core 2 or something? AMD sucks, has for the past few years, please stop trying to make them look better than they are, it's embarrassing!


It shows up in the Far Cry 3 Benchmarks right here on Tom's.

CPU-scaling.png

Farcry 3 is different and its the only game so far to take advantage of more than 1 core. It was designed for tri-core.

But no other games uses more than duo core, let alone the developers using hyperthreading in quad cores.

So farcry 3 as a benchmark is a terrible mistake to make when the general amount of games don't support more than 2 cores.
 

I would beg to differ. If titles are single threaded, then you would think that the FX-4350 at 4.2Ghz would beat the FX-6350 at 3.9GHz. Also, you would expect the i3, which has better single-threaded performance, to beat them both.

Again, I'll refer to these two articles, which has a decent selection of modern titles:
http://www.tomshardware.com/reviews/piledriver-k10-cpu-overclocking,3584.html
http://www.tomshardware.com/reviews/ivy-bridge-wolfdale-yorkfield-comparison,3487.html

If you go game-by-game, and compare the FX-6350 (stock) to the FX-4350 (stock) and i3, I'm seeing that FX-6350 beats both the 4350 and i3 in the following:
Borderlands 2
Crysis 3
F1 2012
Far Cry 3 (Medium Quality)
Hitman: Absolution
Tomb Raider

FX-6350 Loses the following
Far Cry 3 (Ultra Quality)
StarCraft II
Skyrim

So if these games were not optimized for multiple cores/threads, you would see the FX-4350 and i3 beating the 6350 every time, wouldn't you?

In fact, the titles that are the most demanding, where every FPS counts, you want more cores.



 

1. The posts you're replying to are ancient.

2. There are games that are better threaded than Far Cry 3. Crysis 3 is a good example, and it also happens to be more CPU-dependent so it actually matters more.
 
AMD FX-9590 Box is an extremely fast 8-core processor - 4.7 ghz and intel core i5 & i7
speeds are up to 3.8 ghz
What should i choose for gaming, a fst processor, or a really powerful one?
 

You can't compare the clocks, since different CPU architectures don't get the same amount of work done per clock cycle.

The FX-9590 is an abomination and I really can't recommend buying it. Much better to get an FX-8350 or 8320 and save a lot of cash, if you do decide to pick AMD. A Core i5 will tend to be faster in games though.
 


I agree, a 4570 or a 1230v3 would be a far better buy than a 9590.
 


So you have to OC the NB in order to see a 750K or 760K beat the FX Series CPUs? I've always heard a heck of a lot more about OCing the CPU, GPU, and (to a lesser extent) RAM than I have OCing the NB on the MB... Isn't that rather risky and wouldn't you essentially be trying to make the NB simulate being a part of the CPU (the L3 cache and thereby performing a task for which it wasn't built)?
 


You won't necessarily have an FM2 Athlon beat the FX, but it's possible to reach same performance.

Well let's step back and look at what the CPU-NB actually does. It is the memory controller, basically a middleman between the CPU and the RAM (it's actually more comlicated than that but you get the picture). The faster it is, the faster it can relay the data on the RAM back and forth to the CPU. It needs to be at a certain speed to keep up with the speed of the RAM, otherwise the middleman here becomes a bottleneck. Another job of the CPU-NB is that it controls how fast L3 cache runs, actually the L3 runs at CPU-NB speed.

So with platforms that have L3 cache, NB speed is pretty important because it controls L3 cache speed and is the RAM controller.

Platforms without L3 cache, heck it's just as important because you need to have fast RAM to compensate the lack of L3, which means you need a fast NB.

Now, unfortunately due to the way the module architecture is designed, L3 cache performance takes a real hit with BD/PD, and it simply just isn't utilized (with the exception of server loads). See mcdod's quote I made earlier for more info on that. This makes it very easy for a fine tuned Athlon with some fast RAM to catch up with the FX quads. The Richland based 760K in particular has a stronger revised memory controller that makes it easier to run high speed RAM than the slightly older FX chips. (Remember K10.5 C2/C3/E0 revisions, same thing).

Of course, everything here assumes you are fine-tune overclocking everything, and yes ram CPU frequency is paramount.
 


I'm not comparing a 4.3GHz Athlon 750K to a Phenom II x4 at 4.0GHz. But if I was, I would compare the 760K at 4.5GHz to the the Phenom II at 4.0GHz :)

It's at that exact frequency that they both pull 50.0 GFLOPs, and would give you the same gaming performance in older, non-AVX coded games. With AVX, piledriver cores pull ahead quite dramatically. Even without AVX, when Athlon x4 bumps up another notch to 4.7GHz, it beats the Phenom II. If you don't agree, get your 965 out and let's have a bench-off.

 

Hey Damric - what type of cooling and how difficult is it to tweak the 760k to get the performance you're talking about? Is it something you'd expect out of any 760k? I'm just saying I'm interested - have been interested in the 760k but the few 760k reviews haven't shown it was really a worthy replacement of the Phenom II as a budget workhorse that the Phenom II was. It would be great of that level of performance is achievable for the average overclocking enthusiast.
 
To throw more salt on the wound, even a Core 2 Q9550 @ 3.7 is faster than a 750k at 4.3ghz. It is pretty said when AMD cannot beat their own CPU's from over 4yrs ago and much less a CPU from 6yrs ago. I would not consider an FM2/FM2+ chip for a gaming rig, ever. FX 6300 should be the minimum acceptable CPU from AMD, now that Phenom II is discontinued. After the letdown Kaveri was, I cannot, in good conscience, recommend FM2/FM2+.
 


It's be more of a side-grade from a Phenom II x4. Definitley easy to get 4.5GHz though. Much easier then hitting 4.0GHz sustained on a Phenom II x4 as it just doesn't push the same power draw over the motherboard's VRMs. I've seen a lot of of $50 mobos (like A55 chipset) pushing 4.5-4.7GHz. Skim though this thread:

http://www.overclock.net/t/1404334/amd-athlon-x2-340-370-and-athlon-x4-740-750k-760k-be-owners-club

The trouble spot is 5GHz, as there seems to be a hard coded TDP limit, even though no where near a temperature limit. You don't have that problem on the FX chips were I was able to push as hard as I wanted there and was only limited by temperature like the Phenom IIs. We're working on that right now if you read the last page.

 

Well, I don't know how much easier 4.0 GHz on the Phenom II CPU could get.... My 2nd ever build and 1st overclock was a $75 Phenom II X4 965 non-black with a Zalman CLC and $50 uATX MSI Motherboard, and it's been running at 4.1 GHz non-stop for almost 2 years. All it took was setting the multiplier up and a small bump in voltage. Not sure how much easier it gets than that, and this PC runs 24x7.

I'll definitely read through the 760k thread, though. Thanks.
 


Do you have any gaming benchmarks for the 760k running DDR3 2400 RAM? I've tended to agree with you - no good reason to use a 750k. But I've also been reluctant to use anything FX since that is dead-end. Plus, if you want an ITX build, FM2/FM2+ is the only AMD option. Otherwise, you're left with the boring i3, at least for the time being....
 


There was a guy 3 days ago that simply pressed an EZ OC button and was at 4.5GHz stable (post#1205)
 
there were three versions of the PhII x4 965be; only one (c3) was easily overclocked... the c2 stepping, both the 95W and 125W models were seriously capped around 3.7ghz. As for 4ghz or higher on the PhII... why?

The PhII was capped by it's hypertransport/northbridge, not the clock speed. Around 3.8ghz Deneb got almost 95% of it's max performance... sure you could clock it higher, but there was almost no gains in performance. And once you got to 4.0ghz the gains in performance just stopped with deneb. You were almost always better off with a lower clock like 3.8ghz and working like hell on overclocking your northbridge/hypertransport AND ram speeds. better ram timings, faster HT/NB, those things gave you much more performance gains then clocking from 3.8 to 4.0 ghz did with deneb. (note: i'm talking about Deneb, Thuban had a little better performance on the memory controller side of things so clock speeds up to 4.2/4.4ghz could still give you some better performance, but past that point thuban hit a similar wall as deneb did)

That said Deneb AND Thuban were pretty sensitive... it was only late in the runs did you find chips that overclocked easily... frankly, I found the PhII to be a really fun chip to overclock because it was so moody 90% of the time... giving you all sorts of headaches sometimes.



this is sorta true... depends though. When both chips are using the same native instruction sets the PhII will be about 15% faster, clock for clock to a piledriver. Now that's limited of course, as i said earlier, the PhII does not scale linearly with it's overclocks due to a slow memory controller. So up to 3.8-4.0ghz, yes, the PhII will be about 15% faster then Piledriver.

Now then, that's not always the case, the PhII has a limited instruction set it can work with, so there are limited scenarios where it will show such an advantage over piledriver. Furthermore, 15% at 4.0ghz means a piledriver will have to clock up to 4.6ghz to match... which anyone with a piledriver will tell you, is a VERY EASY overclock to achieve... which MOST with a PhII will tell you 4.0ghz is a HARD overclock to achieve (sure a lot of the late run c3 stepping 965be and the 1100Ts could hit 4.0 or higher, but as i pointed out early, atleast for the 965be, there were almost no gains past that point anyway... while the 1100T still saw some diminishing advantage in performance up to around 4.4ghz)

Having moved from a quad core PhII x4 965be (clocked to 4.0ghz) to a FX 8320 I could tell the difference the 4 more cores and advanced instruction sets made with piledriver BEFORE i even overclocked the cpu (same ram, video card and motherboard as the prior system). It wasn't a huge difference but there definitely was an improvement across the board with multitasking, a number of other tasks. Once i had it clocked up past 4.6ghz it was in a different world from that old deneb. The 8 core fx at 5.0ghz is basically indistinguishable from any high end Intel system I've had my hands on (frankly around 4.8ghz i stopped being able to tell the difference)... heck... to tell you the truth, though the fx is a pretty significant step forward in some ways, it's no where near as impressive a gain adding an SSD did for me.

btw: sign me up as an advocate of MOAR CORZ! My early experience with other people's piledrivers (systems i built and the like) kept me from experiencing the advantage more cores can give you like taking a computer i was intimately familiar with (a quad core phII) and replacing a single part (the quad core for a octo core fx) and seeing just how much snappier and quicker everything got on the spot even though rationally i KNOW a fx 8320 at stock is SLOWER a PhIIx4 965be at stock... let alone overclocked. The lone advantage that fx8320 had over the overclocked phII was twice the cores... and damn did it show right off the bat.

Its for that reason, that if i chose to go and get an intel in the future (since it looks like amd has completely abandoned the performance market this year) I refuse to get anything less then an 8 core intel...
 
So basically, if they had just improved the memory controller, added more cores, and was able to achieve higher clocks, Phenom II would have been better. The only real advantage of FX is core count and high clock speeds. I find that kinda sad.
 


Ok, now you are making more sense. Essentially your point is: "IF you overclock every piece of your system that affects CPU performance (CPU, RAM, NB, etc.) as high as they can possibly go and still remain stable THEN (and only then) an AMD 760K can provide equal performance to a [lower end?] FX CPU. Is this a correct statement?

Given this information now, there is one extremely improtant piece of information that needs to be clarified: Are you comparing an overclocked and finely tuned FM2 system to a similarly tuned AM3+ system, or are you comparing an overclocked and finely tuned FM2 system to an AM3+ system with only the CPU or CPU and RAM overclocked?

This HAS to be an apples to apples comparison. If you have no reservations about overclocking the NB on an FM2 system then you similarly won't have any issues doing the same to an AM3+ system. Similarly, if you are not comfortable overclocking the NB on one system then you won't be comfortable doing it on the other.



As a side note:
I get the feeling it has been pretty well established that, in systems with equal RAM and no NB overclock, an FX CPU will beat a 760K CPU. I would venture a guess that 90% percent of users who overclock their systems limit that overclocking to CPU, RAM and GPU, meaning less than 10% of overclockers would even consider overclocking the NB and potentially cutting the longevity of their systems.

IF this information proves to be true then those numbers could change, but I'm sure most people would first want data proving that NB overclocks don't affect dependability or longevity.
 
Status
Not open for further replies.