AMD's Piledriver And K10 CPU Architectures Face Off

Page 2 - 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.

ojas

Distinguished
Feb 25, 2011
2,924
0
20,810
Crysis: Low quality

Surprised that [strike]overclocked processors[/strike] are hampering performance.

EDIt: Ok, it's just one, the 750k.
 


Awesome stuff Paul! Pretty much what I expected, but still great to see, thanks for keeping me in suspense for like a months since telling me about this! LOL curses.

4.2Ghz (turbo) FX4350 being matched by a 965BE at 4GHz. Tut tut tut AMD, well at least my 965BE build is still totally viable after 3-4 years!
 

ojas

Distinguished
Feb 25, 2011
2,924
0
20,810
Nice, informative article. Effectively lays out price/performance leaders in the sub $200 space.

Hope you have something like this lined up for the 200+ segment, plus Haswell.
 

_BigHead_

Honorable
Jun 4, 2013
28
0
10,540
Hey guys! While I do appreciate the absolutely brilliant work you do, with how insanely thorough and in-depth clarifications to things, I can't help but cringe at the use of a custom run when there is a built in benchmark already in a game. Tomb Raider seems like an obvious, taxing benchmark for any good system and I'm sad to see it use a custom run.
Again, thanks for your amazing attention to detail, its appreciated.
 

vertexx

Honorable
Apr 2, 2013
747
1
11,060
Nice article guys! Very informative. I've been looking at the 750k as an alternative to the Phenom II CPU for a budget gaming PC, since the Phenom II has actually increased in price and is starting to be a little harder to come by.

Quick question - is the stock cooler for the 750k good for a mild overclock? Having to buy an aftermarket cooler could make the price/value difference between going with the 750k vs. an I3-3220 build.

Thanks!
 
G

Guest

Guest
So i have this 640 and i ran the memory test on sandra and got 13GB/s...and i'm pretty sure you are using the imc in ganged mode, on both athlon and phenom, which results in lower memory performance
 
I've been looking forward to this article for a couple months now. Now that it's out I'm extremely disapointed that you would leave out the x6 phenom II chips. Why would you do that? If Phenom II has a better per core performance than most FX chips, then it would of seemed logical to include those results. Are you protecting AMD's current architecture/product by not showing what the x6 PH II's are capable of? Then what about an overclocked x6 PH II compared to the FX lineup. Say a 4.0ghz X6 PH II versus a 4.0ghz FX chip, maybe even 4.2ghz. I say 4.0ghz because most topped out around that mark. My x6 1090t hits 4.2ghz, but that probably isn't typical for most x6 PH II's.

Good article with one MAJOR ommision in my opinion.
 

fykusfire

Honorable
Aug 6, 2012
16
0
10,510
Is anyone going to point out the elephant in the room here? From the data, I believe we all can conclude the 6350 and i5 are very close in the benchmarks outside of power consumption. Almost unbelievably close. The article makes mention of that in fact. But what it doesn't make mention of is the price disparity. You can have close to i5 Ivy Bridge performance for 80 dollars cheaper (as of last price check on Newegg). Why was this fact not mentioned Tom's? Don't you believe it is a tad disingenuous not to conclude the 6350 to be the better value considering that alone?
 


Classic case of seeing what you WANT to see. 6350 comes close WHEN OCd. Not to mention a stock I5 can be OCd too...
 


To echo what MariosK said ... your Propus quad might be a dud.

I'm not so sure it will consistently go 13 GB/s on the bandwidth test (that's Deneb territory), but your troubles with NB/IMC scaling are telling. I've typically found the latency 'scaling' improvements of the Athlon IIs beyond that of the PhIIs --- pushing 4-5% for each 10% increase in clock, whereby the Denebs generally go 3-4% (though, it could be a product of my RAMs / timings).

Otherwise ... interesting article. Thanks, THG.

The only thing missing: Popping that OC'ed PhII965BE into something like the Foxconn A79A-S AM2+ 790FX ... :D





 


They had an opportunity to test 6 core vs 6 core the same way they did the quads, and they did not. With games using more than 4 cores now, that information would of been the most interesting aspect of this roundup in my opinion. In the intel article they used a q9550 which was priced more expensive than the PH II x6 1100T chips, so it couldn't of been a price issue of only testing budget chips. I'm not sure why the chip would have to be available though, since a lot of people still run them. I guess it's possible that they couldn't get there hands on one, but seems very unlikely that they wouldn't of been able to dig one up. (Puts on tinfoil hat, lol.) I have a strange feeling that those x6 chips may be better than what AMD would like people to see now that they are no longer selling them. It could be the biggest reason for leaving them out of the article, and is really making me wonder about what the results would of been like in newer games that can use more than 4 cores...
 
Considering the fx6350 is the same cpu as the fx6300 only factory overclocked, i would say THG probably should stop recommending the locked multiplier i5s in their gaming cpu roundup.

Once you branch beyond the same three titles they use to bench the cpus in their roundup the advantages to the 6 core vashera really stand out in a toe to toe fight with a stock i5 (something the building community has been saying for 8 months now). That you can now get the 8 core fx8320 for less then any i5 on the market only highlights what a better budget buy these amd chips are in low budget gaming rigs.

Essentially any gaming rig in the $500-$800 budget range should be built with an AMD cpu.
 


Fully agree with you on the highlighted IF everyone happened to be overclockers...
 

xiinc37

Honorable
Jul 18, 2012
74
0
10,630
On Intels mainstream stuff (socket 115x) IOMMU is disabled on chips/chipsets with unlocked multipliers. (You can get around this by going with socket 2011). With AMDs stuff you can have IOMMU and an unlocked multiplier without spending $600 on a cpu. So if you do virtualization (linux host daily driver - windows guest for games) thats a good thing. Just thought I'd throw that out there.
 

salgado18

Distinguished
Feb 12, 2007
980
437
19,370
Oh, just get over it already. Phenom II's have a better IPC than FX's, everyone knows that. But look at the benches, Piledriver has a much higher clock than the older architectures, even at stock settings, and better performance as a result. Yes, it is a bit less efficient clock-for-clock, but it compensates with a much higher clock at similar wattages, so it is better.

This discussion came up when Bulldozer came, and it made sense back then, with clocks in the 3.6GHz range (similar to the 965BE), but now they are mostly around or above 4.0GHz, which is the sensible limit of a Phenom. "But a Phenom can reach 4.2 on air!" And an FX-4350 reaches 4.7.

Piledriver has become better, Steamroller will be even more, and old processors are old. Just move on, please.
 
I'd like to see more in depth exactly what causes some of the performance differences on tests and why some CPUs do better on some tests than others. It would be nice to see a list of features to turn down for each architecture in order to optimize the performance depending on the CPU.

For instance, take a look at Tomb Raider, which is mostly regarded as GPU dependent. On the chasm test on both high and ultra presets the stock 6350 just edges out a stock 3570k (by about 4% on average), yet on the Mountain test the 3570k beats the AMD on the minimum frame rate by about 40%. So what exactly is causing that discrepancy.

The same type of interesting results on the Crysis 3 test where on both Low and Medium the 6350 wins, by about 15-25% depending on the test, but then on High the 3570k wins by about 8% on the average and 24% on the minimum.

So which settings cause these types changes and why, it might make for good information to have when trying to get the maximum performance out of your games depending on what architecture you are running.
 


not all phenoms are the same. the 1100T is almost 15% faster then piledriver at the same clock. that's a significant difference. No matter how you look at it, bulldozer/piledriver were a step backwards. If steamroller can make the 20-30% IPC improvement they're talking about then we can say it's an improvement over the k10 architecture.
 


It says nothing about it IMPROVING your overclocking, or even allowing better overclocking. It says nothing about powersaving features either... It says that BCLK overclocks are functioning AND reporting oddly in windows8, and because of this they no longer will accept win8 benching results. Interestingly, AMD cpus have no such issue in windows 8.

Basically it's saying that Windows 8, yet again, falls short for the enthusiast community.
 
Status
Not open for further replies.