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

Page 3 - 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.
@UltimateDeep; if you already are spending obscenely more than you need to, there is no saying "I wouldn't use the third SLI capability anyways"

Then don't go 2011-3! There's no reason to. If you reply soundcard, i reply "learn to audiophile", if you reply PCIssd, I reply "raid0 is faster"

If you're making compromises, then you shouldn't be getting 2011-3 anyways. If you're not making compromises, there are benefits to the 5930k. If a rebuttal is then why not 5960X? Because then there are >no< tangible benefits to the massive cost increase. Features for gamers stop at the 5930k, so that's where the recommendations stop too. :) Hope I cleared that up, sorry if I sound a little prickish!
 
"For now, our recommendations are based on stock clock speeds and performance at that price. " And then for the reviews it mentions overclocking as a big draw...odd.
 
The hierarchy chart needs some work/adjustments...

The top tier of the chart has a 50-100% performance discrepancy from worst to best chips represented for real-time workloads.

Then there are chips in the tiers below, separated by 2 tiers, that have less than a 30% difference in performance.

There are also several chips in the top tier that would be out-performed by some of the chips in tier 2 in gaming overall.

I'd like a go at "fixing" the chart. Anyone interested in that?
 
As has been said so many times here, these charts are for gaming performance.
Hence, the CPU ordering is perfectly ok. Infact if anything it's on the conservative
side, eg. right now I'm benchmarking an i5 760 setup with 7970/CF. For GPU-bound
tasks, even at stock speed the system is only 1% slower than the same cards running
with a 5GHz 2700K, while for mixed-loads the deficit is nothing like as bad as one
might imagine, especially once the CPU is oc'd (easily outpaces a stock 870).

Of course there are large differences between the CPU tiers for CPU-based tasks,
but this is not a CPU performance chart, it's a gaming chart, and this means
what is adequate can vary in all sorts of unexpected ways (in the case of P55,
a lower latency platform helps).

Ian.

 
I am talking about gaming/real-time workload performance. The CPU ordering is certainly NOT perfectly OK, there are some glaring flaws.

An FX-9590 is faster than an i7-975 in these workloads, why is the 975 in the top tier while the 9590 is in the same tier with an FX-4350 (which, FYI: has basically the same performance in real time workloads as the i5-2300 found in the tier above).

If all CPUs perform the same in gaming, then there would be no point in having the chart or the monthly article. Obviously, in the real world, there are compute intensive games that can not be bench marked because conditions are not repeatable. In the real world, there is performance scaling with compute performance. Real time workloads still strongly gravitate towards the availability of execution throughput to 2-4 threads. There are CPUs in the top tier with performance in these types of workloads that is worse than some of the CPUs in the second tier. I believe the chart is very flawed right now.
 
In the 'real world', as gaming shifts towards higher resolutions, reliance on CPU
power is reducing anyway, and the differences are not that large IMO. Besides,
anyone who chooses a 9590 over a 975 or anything else is crazy. 😀

Like I say, people would assume there would be huge differences between
an i5 760 and a 2700K, but there really aren't in most cases, especially at
high res/detail.

Ian.

 
In the 'real world', as gaming shifts towards higher resolutions, reliance on CPU
power is reducing anyway,

This is false. Compute workload does not go DOWN with increased resolution. The only time the compute workload is reduced is if the FPS is reduced. If you want 60FPS in a compute intensive game, it doesn't matter if you're playing at 720P or 4K, the compute requirements certainly will NOT be LOWER at 4K than at 720P for equal FPS. The compute workload is always at least the same or slightly higher to maintain the same goal FPS at a higher resolution.

---------

Furthermore, the practicality of the FX-9590 is irrelevant within such a chart. If the title of the chart were "most practical CPUs for gaming" then fine, put the FX-9590 down with a Pentium 4 for all I care and put the i5-4590 above the rest, but that's not the purpose of the chart. The purpose is to categorize performance classes for gaming workloads, something that it currently fails to do so with reasonable accuracy. The execution throughput of the FX-9590 for real-time workloads is as high or higher (depending on the specific workload) than the i7-975 no matter how you slice it and no matter how impractical the 9590 may be. Any chart which places the i7-975 in a tier ABOVE the FX-9590 has some serious BIAS issues, and this is coming from someone who has spent (likely wasted) hundreds of hours of my time researching and explaining the advantages of current intel CPU architectures for real time workloads. The i7-975 is bloomfield, a 4 generation old CPU. The execution throughput available to any single thread on the i7-975 is actually very similar to PileDriver chips with Turbo speeds in the low 4ghz range. The combined execution throughput available on all 4 cores, when saturated with an 8 threaded workload, is no better than the combined execution throughput of an FX-6300.

http://www.cpubenchmark.net/singleThread.html
http://www.cpubenchmark.net/high_end_cpus.html

Note: the per-core performance of the 975 is comparable to PileDriver chips (+/- chips like the A8-6600K, 760K, FX-4350/6350/8350, etc). The combined throughput lines up with the FX-6300. I fail to see how this chip deserves to be in a "class" for gaming performance (which scales first and foremost with per-core performance) above the FX-9590, when it offers little to no compute advantages over a 6 core piledriver.

AMD originally slated bulldozer to compete with Bloomfield, and PileDriver to compete with Gulftown, obviously they were late to the party with those releases, but it's important to understand that Bulldozer and Piledriver architecture is technologically matched to that same erra, and does compete with intel products from that erra very well. No surprise then that the FX-8350 offers execution performance in both lightly and heavily threaded workloads that is very comparable to an i7-970 (gulftown 6 core).
 

If you virtualize and need moar cores to run your VMs, a $140 FX 8320 hax AMD-V and lets you run 7 different VMs with a core each, rather than $1K for an Intel 8-core. Also, if you use linux more than Windows, the performance differences are much smaller due to better parallelization of linux - FX-8350 regularly trades blows with i7-4770K rather than i5-4430 (phoronix). Some of us also live way the hell up North, and a spaceheater is welcome for 8 months of the year. So, there are reasons.
 
As a Linux user myself, I am very happy with the AMD platforms. AMD-V, ECC support, and IOMMU's all on dirt cheap hardware. There are some very real advantages if you're into poking and prodding and learning enterprise systems without an investment in enterprise hardware.
 


That depends on the game. Actually my point was more that for GPU-bound titles, the CPU factor is less
of an issue at higher resolutions & detail. For CPU-affected games, it very definitely does vary. I've seen
the effects quite clearly with my own tests. And some games are totally CPU-bound, like X3TC.

Ian.

 
i5 4690k stock is A LOT around 30% slower than my old q9650 stock I have tested MAY BE i7 4790k stock is equal may be so I m happy with my cpu.
 
The newly launched Athlon X4 860K didn't even get a spot in the hierarchy charts. Personally I think it would have put up a good competition to the i3 4130 in the price to performance segment.
 
Can't believe this list isn't all over the 5820K which has to be the performance sweetheart. . Just swapped into my desktop a 5820K, MB and 16g 2133 DDR4 for a total of $727 before tax. The additional $600 for a 5960K would cover one GTX 970 and an H100i clc and half of a second GTX 970 for SLI.
 


Because it isn't a substantial improvement, over an i7 4790k, for games. You can get a 4790k setup for under $600. For straight up gaming, you really don't need anything beyond an i5 4690k.

PCPartPicker part list / Price breakdown by merchant

CPU: Intel Core i7-4790K 4.0GHz Quad-Core Processor ($309.99 @ Amazon)
Motherboard: Gigabyte GA-Z97X-Gaming 5 ATX LGA1150 Motherboard ($144.99 @ Amazon)
Memory: G.Skill Ripjaws X Series 16GB (2 x 8GB) DDR3-1866 Memory ($144.00 @ Newegg)
Total: $598.98
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2014-10-01 15:24 EDT-0400



PCPartPicker part list / Price breakdown by merchant

CPU: Intel Core i5-4690K 3.5GHz Quad-Core Processor ($225.98 @ SuperBiiz)
Motherboard: Gigabyte GA-Z97X-Gaming 5 ATX LGA1150 Motherboard ($144.99 @ Amazon)
Memory: G.Skill Ripjaws X Series 16GB (2 x 8GB) DDR3-1866 Memory ($144.00 @ Newegg)
Total: $514.97
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2014-10-01 15:26 EDT-0400
 
Anoone noticed that games are so f****** old. Skyrim is technically year 2006, Starcraft 2010 and Far Cry 3 2012. It's 2014 now.

That's why dual core seem to be fast, however on every modern game at least quad core is recommended.
 
You guys kind of miss the point. Any processor above a 4430 is going to perform nearly identical in GAMING all the way up to the 4690K. Everything after that is going to perform nearly the same. Basically current CPU above 4690k will not bottleneck the current GPUs on the market.anytime soon. Most of the next generation games are mostly GPU dependent. The only games that are highly CPU dependent are really old games that still use single threaded CPU performance. Higher ghz on a single core.. Those games will benefit from a faster core. Many of the new games are multi thread and barely take full advantage of quad core.
 
It's a good prediction that new i7 and AMD 8 core cpu's will gain on gaming worth in next years. If we are talking about getting a new computer then a timespawn until 2018-2020 is realistic for modest buyers (who didn't buy a 500$ i7) and PC games simply should have picked up the 8-thread trend from XBOX and PS until then.
 
I think we need to straighten this out - AMD have never released a processor with SMT. Yes, it's possible that they're going to release such a processor architecture in the future, but we've not seen such products as of yet.
 
I'd like to see AMD's 25W Kabinis added to the CPU chart. PCPER suggested in an older article that the chip is viable, even with its 4-lane PCIe limitation. So, where does it fall, please?
(...and no, "on its face" is not a useful answer)
 
Status
Not open for further replies.