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

Page 36 - 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.
I have a doubt when looking at this chart. FX 6300 cant be such bad value , even being worse than an I3. Could it be due to the fact that the average is comprised of just 3 games , 2 of them favouring strong single core performance?
 

Games strongly favoring single/dual-core performance represent something like 99% of titles out there so, while the game mix might not allow massively threaded CPUs to shine, it is a fair enough representation of what is out there when you look at games in general rather than only the highest of high-end titles.
 


I had this discussion a while ago with the THG staff. I think the idea is to be representative since they can't just show performance in every game. personally i think 3 titles is a questionable metric, but my suggested fix probably would have been even more a lopsided win for intel... as i was suggesting they make the games more representative to the "average" game.

My main objection was to the fact 1 of the titles use x87 code, which is really only used in like 4 games... that said the staff had an excellent counterpoint, which is 2 of those 4 games are top 5 selling titles for the last 3 years. There is no argument against this.

I think the only part of my argument they took to heart and promised to discuss was swapping out the "almost" fully threaded Far Cry 3 (its mostly designed to run on 4 cores; it will use more, but it won't use them much, as most of it's processes will be dedicated to the first 3 or so cores) for a completely and truly fully threaded title like BF4 or Crysis3 which will give people a little more idea of what to expect from next gen console imports.

for now if you want to know what to expect from MOAR CORZ! you can look at the FC3 results... they aren't great but they'll show you a more representative result then the other two titles will...
 
With the multitudes of data THG has with recent reviews comparing Core 2 to Ivy and K10 to piledriver, they should be able to properly fix the charts in a way that they make more reasonable sense. Those two articles were great reads, with lots of information. Sadly, they also tend to contradict the hierarchy charts. I don't expect perfection in the charts, but improvement can be made and the base is there to make the charts more accurate. Adjusting the charts to reflect the data in those two articles would clear up quite a bit of confusion. Having a PhII quad above an i7 is just plain silly. :lol:
 

This has been my point for the last few months ( x87 code aside. ) Yes, games are slowly starting to utilize more cores, but for now the vast majority are single-threaded. A few of the bigger titles will use an extra core or two, but very few comparatively. With this being the case, single-threaded performance needs to be a big priority in scoring a CPU, at least for the near future.

Steam stats show a 50/50 split between dual- and quad-core CPUs. That hasn't changed in the last five months ( I wish I could find a historical trend farther back, if anyone has one, please let me know. ) Game devs still want their games playable by as many people as possible, so they're not going to start optimizing for quad-core and beyond until that becomes the vast majority of their consumer base, probably somewhere around 66% - 75%.

The PS4 and XB1 might speed that along a bit, but I'm willing to bet that it'll be at least two years before we start seeing mainstream games that want four cores. By then we're talking completely new architectures from both AMD and Intel that may well make this argument moot.
 

AMD's push for HSA/hUMA/GPGPU might make extra CPU cores superfluous for gaming if successful with most compute-intensive stuff shifted off-core and I would bet Intel has their own implementation coming up, maybe for Skylake in 2015.

It does not make much sense to run highly linear compute-intensive code on cores optimized for relatively non-linear code when resources for embarrassingly parallel execution are available so if there is going to be a paradigm shift in parallel programming for mainstream applications and games, might as well go the extra mile to run all the massively parallel stuff on hardware that actually has the potential to scale massively.
 

Hmm, really good point. Question is, if GPU resources are reassigned, how many gamers then start complaining that graphics fidelity starts to suffer. I can see it now, some people just can't be appeased.
 

Why would there be a loss there? People with discrete GPUs for gaming aren't using their IGP for graphics, which leaves it available to run the math-intensive parts of physics, DSP, AI and other algorithms.
 

IGPs aren't the only potential target for offloading, and not all CPUs have IGPs either.
 

They may not all have IGPs but which current mainstream CPUs lack IGP? Pretty much only AMD's FX series. What are AMD's plans for the FX series on their 2014 roadmap? None. If you read the writeups about AMD's presentations last week, AMD will be betting the farm on APUs and heterogeneous computing for the foreseeable future.

By the time hybrid computing catches on, AMD will be on 20nm or better with hybrid computing benefits proven well enough to cause people with FX chips to itch for an upgrade to APU.

It is just another example of chicken-and-egg problem: nobody will write for hybrid computing if there is no hardware supporting it and nobody wants to make the hardware unless there is software that will actually use it. Once the ball gets rolling, the rest sorts itself out - people who lack an IGP good enough for hybrid computing will upgrade if the perceived benefits are worth it. Not much different from any regular CPU upgrade.

PC manufacturers are somewhat desperate for a new application that may whip up PC sales. The transition to hybrid computing could be one of those things that deliver enough benefits to make people upgrade.
 

Sorry, wasn't trying to spark an argument, only trying ( and apparently failing, ) some sarcasm for what some graphics fanboys will inevitably complain about when some processes are offloaded to their GPU.

Right now the Intel P series, AMD 750K and 760K, and half the E3 Xeons don't have IGPs ( I'd argue the E3 1220 and 1230 could be mainstream while the upper SKUs certainly are not. ) And of course the FX and earlier Thuban and Deneb chips that are still rocking away. But as you say, Intel shows no sign of removing the IGP, and AMD definitely is promoting their APUs above everything else.

I do find your proposed scenario intriguing, but I don't think it's quite as "chicken-egg" as you seem to. I see it as slightly more likely to happen than not. It reminds me a bit of early math co-processors, except this doesn't take additional silicon. And unlike PhysX, this is open to all without licensing fees and vendor specific hardware. Basically the hardware to support this is already readily available ( provided Intel's IGPs support the OpenCL calls, and I don't see any reason they'd lock them out. ) The only cost of entry is coding for it.
 


You should grab another one then as a back up at microcenter XD they are on sale for 149.99. Now is the time to get a 3570k if someone needs or wants one.
 


Wow $149.99 for an i5-3570K is a killer deal.
 


Yep if I had the money i'd get it and just hold onto it. Because its not going to retail for that low for a couple years at least. Only way you would find one cheaper is used.
 


As fast as in clock speed? You'll be waiting a long time.
 


:'( :'( :'( "available only for pickup in store" :'( :'( :'(

EDIT: OH $hit they got a store in philly!!!! (I'm minutes from the bridge)
 

Actually people that want to buy a new CPU would be like:
3ru4u8.jpg
 
Memnarchon I didn't run fast enough :'( out of stock

Secondly why the hell doesn't this stuff show up in google or bing? I've been checking like every 24-36hours for crazy deals like this and coming up empty, just happen to check this thread and boom, much cheap processoro
 

Probably because the sales do not last long enough to get caught by web-crawlers.

With the number of people checking sites to respond to posts on THG forums and elsewhere, they collectively probably have much higher chance of catching sales nearly as soon as they get posted than search engines do.
 


I found out through pcpartpicker.com while I was putting together some builds I saw 149.99 and knew that wouldn't last long.
 
Status
Not open for further replies.