
As you already know, we run two scripted Photoshop benchmarks: one that is threaded and runs on our CPUs and another that leverages OpenCL acceleration to tax our graphics cards. They're completely different workloads, which is why performance varies so much.
In the CPU-oriented test, last quarter's FX comes out ahead at stock clock rates. Then, both processors tie when we overclock. The OpenCL-based benchmark shows a huge advantage favoring AMD's Radeon HD 7870 over the GeForce GTX 670.

Meanwhile, we know that Adobe's Acrobat X is single-threaded. Naturally, then, our PDF document creation workload favors the Core i5 stock and overclocked.

Premiere Pro's media encoder is threaded, and our results from that benchmark consequently look a lot like the other content creation apps we've already tested.

After Effects has always seemed very sensitive to memory allocation, and AMD's ability to operate on eight threads in parallel might actually be hurting it in our AE benchmark. The Intel platform gets more memory per core, and consequently turns in notably better scores at stock and overclocked settings.

- Building A PC: What Do We Get For $800?
- CPU, Motherboard, And Cooler
- Video Card, Power Supply, And Case
- Memory, Hard Drives, And Optical Drive
- System Assembly And Overclocking
- Test System And Benchmarks
- Results: Synthetics
- Results: Media Encoding
- Results: Rendering And Productivity
- Results: Adobe Creative Suite
- Results: Compression Tools
- Results: Battlefield 3 And The Elder Scrolls V: Skyrim
- Results: F1 2012 And Far Cry 3
- Power And Temperature
- When Does $800 Buy You More Than $1000?
And therein lies the problem with benchmarks.
An enthusiast PC, without a SSD boot drive?
Overclocking the Core i5-3550K is"
there would be marginal performance boost from switching from a 7870 LE(nerfed 7950, heck can call it a 7930 and it would be partially correct in a way) to an actual 7950. Though its likely the outcome for the 1k budget coming up next.
That's cuz the 8350 is using a 670 which in GPU heavy titles will boost its numbers higher. Same GPU would show a more different story and the price difference between a 8350 and a i5 3570k is only able to bump a 7870xt to a 7950 at most, not to a 670
And therein lies the problem with benchmarks.
An enthusiast PC, without a SSD boot drive?
For a $800 budget, I would rather get all the real in game performance I can first while and add a ssd later than lose out on fps and get faster load times
higher fps(stronger cpu, gpu) > faster load times
$35 saved from cutting cooler and k is not enough for an SSD
After looking at this it would seem illogical to buy a 8350 over a I5. But yes it does do decent interns of price/performance.
This would have been correct for a "$800 Gaming PC" .
But for a "$800 Enthusiast PC " , a SSD is a must. Even a 64GB, lower end SSD would have been OK.
this build looks like a budget-upper-midrange build (if that makes any sense). the mobo... looks weak. the cooler and gfx card looked... cheap. i didn't expect the oc core i5 3570k build to keep up with oc fx8350 build in threaded benches (for $200 less, even). only 7zip seems to take advantage of 8 integer clusters/cores properly and the rest of them don't seem to scale well beyond 4~ cores. i noticed that trend in games but this is the first time i've seen it in non-games softwares. i use handbrake, lame mp3 and archivers (7z, zip/rars), so those benches were very informative for me. thank you.
when i first started reading, i wanted to see an fx8320, cm hyper 212 evo(or a corsair clc) with a sturdy 970 mobo + radeon 7870xt. as i read on, this current build and its performance started to look more and more interesting.
The difference between the 7870 XT and the 7950 can be huge when overclocking is considered. That lost memory bandwidth is no small matter for Tahiti LE when it runs at around 1.2GHz. I also suspect that the lost compute units from 28 to 24, although not a significant loss, are considerable.
EDIT:
http://www.tweaktown.com/reviews/5120/powercolor_pcs_radeon_hd_7870_tahiti_le_2gb_myst_video_card_review/index7.html
Also, not only are the 2560x1600 bencmarks not the only ones showing such a comparison, but also the 1920x1200 and 1680x1050 benchmarks. This is also fairly consistent across most games. Both of these claims are demonstrated to be true by the rest of this article.
For a $800 budget, I would rather get all the real in game performance I can first while and add a ssd later than lose out on fps and get faster load times
higher fps(stronger cpu, gpu) > faster load times
$35 saved from cutting cooler and k is not enough for an SSD
We could easily scrape out enough money from the budget for a decent 60GB/64GB SSD such as Plextor's M5S 64GB without really hurting core performance, at least if we didn't stick to Nweegg (granted Tom's doesn't have much option left in that if they want free systems to hand out). Loading times alone could be worth it.
I'm wondering if Blaz's disable-one-core-per-module trick would help Piledriver here, as you'd have a single core with access to 2MB L2 and 2MB L3 without the scheduler needing to worry about the second integer core. As it is, even if the software could make full use of all the CPU cores, they'd likely have a memory contention or bandwidth issue.
Steamroller will definitely improve matters but AMD will continue to be behind until anybody but the creators of 7Zip thread their software to hell (outside of rendering and productivity apps, of course).