HardOCP: FX8150 vs. 2500K with 2-way/3-way SLI

[apoppin]

You really need to upgrade your Intel setup, it is ancient. The HardOCP review mentions that they got up to 40***37; higher FPS in multi-GPU tests upgrading from the 920 to the 2500K.

i want a real upgrade - not 2600K's minimal improvement for gaming - i don't care about encoding ... and my current i7@4.0GHz is plenty fast to demonstrate scaling with GTX 580 SLI and HD 6970 Tri-Fire

What is ancient about i7@4.0GHz? Until last month, HardOCP used the same CPU that i do. X58 is still Intel's flagship platform (for a few more days) and my i7-920 does 4.0GHz and even higher - Intel's i7-960 is current and it only runs at 3.2GHz

My CPU is easily the equal of a stock i5-2500K in gaming. And it has no issue with scaling with Tri-Fire. That said, IF it doesn't scale with Quad-Fire-X4 6970s, then i will get a faster CPU - probably Sandy Bridge-E from Intel for evaluation. i am evaluating i3 right now.
:whiste:

 

[HopJokey]

i want a real upgrade - not 2600K's minimal improvement for gaming - i don't care about encoding ... and my current i7@4.0GHz is plenty fast to demonstrate scaling with GTX 580 SLI and HD 6970 Tri-Fire

What is ancient about i7@4.0GHz? Until last month, HardOCP used the same CPU that i do. X58 is still Intel's flagship platform (for a few more days) and my i7-920 does 4.0GHz and even higher - Intel's i7-960 is current and it only runs at 3.2GHz

My CPU is easily the equal of a stock i5-2500K in gaming. And it has no issue with scaling with Tri-Fire. That said, IF it doesn't scale with Quad-Fire-X4 6970s, then i will get a faster CPU - probably Sandy Bridge-E from Intel for evaluation. i am evaluating i3 right now.
:whiste:

Ahh yes for personal use your Nehalem is decent now, however it seems odd benching the latest from AMD versus a product from Intel that came out in 2008.

If you can get a 920 @ 4 GHz you can definitely get the 2500/2600K to 4.8 GHz maybe even 5.

 

[Skurge]

Im pretty sure my old phenom2 X4 would do better than that.

Seriously though. BD better kick ass on the server side or AMD have no idea what they are doing.

I find it hard to believe this is the same company that gave us Athlon. Phenom1 wasn't even that bad. A Die shrink fixed it. I don't think a shrink would fix BD for gaming at all. I can see it helping with other apps, just wouldn't touch or recommend it for gaming at all.

Was such a huge AMD fan. At least their GPU division and Fusion is doing good.

 

[apoppin]

Ahh yes for personal use your Nehalem is decent now, however it seems odd benching the latest from AMD versus a product from Intel that came out in 2008.

If you can get a 920 @ 4 GHz you can definitely get the 2500/2600K to 4.8 GHz maybe even 5.

i think it kinda makes a point, doesn't it?
:\

As soon as i see my i7 start to slow down QuadFire, i will upgrade, no worries. ... as a practical need, there is NONE at the moment .. except to "keep up" with the other tech sites.
()

 

[mmaestro]

Yeah, the frame rate is all over the place in some games, I don't really understand that. Does the same happen with Phenom II? If not maybe it has something to do with the scheduling issues in Win 7 for BD. For example if the OS schedules 2 threads to the same module one second and 2 threads to different modules the next, that could result in a pretty large swing in performance and might explain the variance.

This had been my thought, too. I would love to see the same bench run on Windows 8 just to see how it comes out. IMO, the biggest mistake AMD made may well have been not going to Microsoft on bended knee with some cash to write a patch to fix the scheduling in Windows 7. We don't know how big a difference it would have made in the end, but I suspect that it would make Bulldozer just a mild disappointment, not the train wreck we see here. Of course, without benchmarks using the right scheduler, that's just speculation.

 

[3DVagabond]

Very interesting review there. Pretty rare as not every reviewer has 3 GTX580.

http://www.hardocp.com/article/2011/11/03/amd_fx8150_multigpu_gameplay_performance_review/1

Not really sure what to think about AMD promoting their FX platform for (highres) gaming...kinda backfired. Badly. Anyway, enjoy reading.

They did it trying to hide BD's warts. At hires's the bottleneck shifts to the GPU. Give it enough GPU power though and the bottleneck will shift back to the CPU.

 

[3DVagabond]

This is not news, is it ?
We have all been saying that AMD's new flagship is just too expensive for what it brings to the table.
This subject has been beaten to death, yet people still keep saying the same things over and over.
I know, they should have used crossfire to make it better!

 

[frozentundra123456]

This had been my thought, too. I would love to see the same bench run on Windows 8 just to see how it comes out. IMO, the biggest mistake AMD made may well have been not going to Microsoft on bended knee with some cash to write a patch to fix the scheduling in Windows 7. We don't know how big a difference it would have made in the end, but I suspect that it would make Bulldozer just a mild disappointment, not the train wreck we see here. Of course, without benchmarks using the right scheduler, that's just speculation.

The last hope for Bulldozer fans: "well it is ahead of its time, it will work in Win 8".

Why one would design a CPU for an OS that is not even out yet is beyond me. And if Bulldozer had come out on time, it would have been even longer before win 8 came out. But as you said, we dont even know if performance will be significantly improved in Win 8.

 

[frostedflakes]

Win 8 Developer Preview shows up to 10-12% more performance in some games from what I've seen. It does look like the scheduler tweaks can offer some pretty solid gains when running four threads or less.

Don't get me wrong, I wouldn't buy a BD now and wouldn't recommend most people buy one now either. 2500K in general seems to be the better CPU if you want something that offers great performance *right now* as opposed maybe better performance 6-12 months from now. But you do have to keep in mind that BD is a completely new architecture and will probably see much better performance once software is optimized for it. The sub-optimal thread scheduling is one thing that really seems to hurt single/lightly threaded performance in Win 7. Just like a HyperThreaded CPU would perform like crap in an OS that isn't optimized for it and isn't aware of the difference between a physical core and a logical core, BD performs kind of crappy in an OS that isn't module aware and doesn't know how to optimally distribute threads. The difference is that people have had a decade to optimize software for HyperThreading, whereas BD and CMT is relatively new. It's not wishful thinking or anything like that, it's just the reality of a completely new architecture like BD. Intel experienced the same growing pains with their SMT scheme. Things will probably improve over the next year or so as software is optimized for it. But as mentioned, that doesn't change the fact that it's a poor choice right now. But I'm not willing to call the architecture itself a failure yet, with more optimized software it could be more competitive.

 

[Hypertag]

Win 8 Developer Preview shows up to 10-12% more performance in some games from what I've seen. It does look like the scheduler tweaks can offer some pretty solid gains when running four threads or less.

The difference is that people have had a decade to optimize software for HyperThreading, whereas BD and CMT is relatively new.

Yes, this article proves that with just 12% more performance that bulldozer would be competitive, and the only reason that bulldozer lost is that Intel paid game developers to take advantage of the 2500k's hyperthreading.

 

[frostedflakes]

Never claimed anything remotely like that.

Just saying that software optimizations will help. Scheduler improvements look like they'll bring IPC of Bulldozer in single/lightly threaded situations more in line with Phenom II. That (along with power consumption) is one of the areas I was really disappointed with Bulldozer, so it will be nice to see improvements there. If optimizations like that can make it so BD can offer comparable IPC to Phenom II, plus the solid heavily threaded performance it has, IMO it will be a much better option and more like what people were expecting BD to offer before release.

 

[Seero]

Yes, this article proves that with just 12% more performance that bulldozer would be competitive, and the only reason that bulldozer lost is that Intel paid game developers to take advantage of the 2500k's hyperthreading.

Nvidia paid devs to utilize their GPU better, and Intel paid devs to utilize their CPU better. AMD ...

In case you don't know, 2500k have 4 physical cores and no HT. The one that has HT is called 2600.

 

[RussianSensation]

Yes, this article proves that with just 12% more performance that bulldozer would be competitive

Bulldozer ships at 3.9/4.2ghz Turbo and even at 4.8ghz overclock, it didn't do much to alleviate the CPU bottleneck in games. So no, a 12% performance increase over FX-8150's stock speeds, or even over 4.8ghz won't do much at all. Right now, Sandy Bridge has about a 45-50% advantage in instructions per clock vs. Bulldozer. Since most programs don't really use more than 4 threads (specifically games in this case), a 4.8ghz 2500k ~ 7.2ghz FX-8150. It's no wonder that the FX-8150 4.8ghz got smoked in that review.

, and the only reason that bulldozer lost is that Intel paid game developers to take advantage of the 2500k's hyperthreading.

2500k doesn't have hyper-threading...it's just a quad-core.

The reason BD lost to SB is because on a per core basis (even without Hyper-Threading), SB is much faster at the same clock speed:

Source

 

[busydude]

The reason BD lost to SB is because on a per core basis (even without Hyper-Threading), SB is much faster at the same clock speed:

There is a definite improvement between 2M/4C Vs 4M/4C.. that test uses 2 cores from the same module, but AMD says Win 8 will allocate threads with minimal loos in performance. The performance loss when both cores in a module is anywhere between 30%-70% IIRC.

Even then it loses pretty miserably to SB though.

 

[JoJoman88]

That would be unfair to AMD. AMD's actual competition from Intel is http://www.newegg.com/Product/Produc...82E16819116404 , and anyone that doesn't bench the FX-8150 against that chip is an intel employee.

Wow 84 watts for a single core, talk about space heaters indeed! How about a 336 watt quad core to heat your case

 

[Silverforce11]

BD is such a rubbish CPU. I've ordered a 2500k/z68mb for cheap and OC it for my new rig. Hopefully IvB is compatible with z68.

 

[Qbah]

Yes, this article proves that with just 12% more performance that bulldozer would be competitive, and the only reason that bulldozer lost is that Intel paid game developers to take advantage of the 2500k's hyperthreading.

Looks like people missed your sarcasm :awe:

 

[zlejedi]

The last hope for Bulldozer fans: "well it is ahead of its time, it will work in Win 8".

Why one would design a CPU for an OS that is not even out yet is beyond me. And if Bulldozer had come out on time, it would have been even longer before win 8 came out. But as you said, we dont even know if performance will be significantly improved in Win 8.

It must that famous price per performance AMD always brings to the market - you only need 250$ cpu, 200$ new OS and expensive ddr3 1866 memory to even hope of matching that overpriced $220 2500K

 

[StrangerGuy]

It must that famous price per performance AMD always brings to the market - you only need 250$ cpu, 200$ new OS and expensive ddr3 1866 memory to even hope of matching that overpriced $220 2500K

Forget those, they can't even win it conclusively with a $110 CPU.

 

[jacktesterson]

I just built my 2500k system.

I really wanted a bulldozer system, but it failed.

 

[exar333]

It must that famous price per performance AMD always brings to the market - you only need 250$ cpu, 200$ new OS and expensive ddr3 1866 memory to even hope of matching that overpriced $220 2500K

Please explain how a $250+ BD that is slower than a $220 2500k is a better buy? 2500k is faster and uses less energy.

Raw performance: Intel wins
Perf/$: Intel wins
Perf/w: Intel wins
Power consumption: Intel wins
OC percentage: Intel wins
Perf scaling w/ OC: Intel wins
Better CPU box/packaging: BD wins (cool little tin...)

I threw AMD a bone here on the last one.

 

[Idontcare]

I just built my 2500k system.

I really wanted a bulldozer system, but it failed.

You'll be sorry when you find yourself in the middle of a winrar battle with your p0rn archives

 

[boxleitnerb]

Bulldozer might come in handy when you forget that password to all the p0rn

 

[blckgrffn]

Please explain how a $250+ BD that is slower than a $220 2500k is a better buy? 2500k is faster and uses less energy.

Raw performance: Intel wins
Perf/$: Intel wins
Perf/w: Intel wins
Power consumption: Intel wins
OC percentage: Intel wins
Perf scaling w/ OC: Intel wins
Better CPU box/packaging: BD wins (cool little tin...)

I threw AMD a bone here on the last one.

*Taps Exar's sarcasm meter - yep, it's definitely broken...*

 

[evident]

I know huh.

The funny thing is there are alot of people out there who think its not that bad, and that the BD sucks threads are to negative and are just people jumping on the anti BD bandwagon and that its just a mid range chip thats ok. When the cold hard truth is that IT BLOWS, its really that bad, the threads are justified

lol, like i said in the other thread, this cpu is trash no matter how you put it.