Give me a reason to go AMD... I WANT TO BELIEVE.

[Gikaseixas]

After hearing things like amazing performance from the new scheduler I'd like to know too.

Buy one, test it

 

[Insert_Nickname]

I'm not surprised either. I think Win8 runs better than Win7 on either CMT or HT based processors. I ntoiced Win8 runs really nicely on my 3770k too. Haven't installed it on my FX8350 but I'm glad to hear from the sounds of it I will see a nice improvement on the desktop PC there as well.

Interesting. I was going to install Win8 Pro on my main system. Perhaps I should move that forward a bit, to try this out...

This is the standard Intel HT behavior of the Windows 7 cpu scheduler on the FX, all modules are active and not power gated with light and medium workloads using all available module resources, turbo core doesnt kick in aggresivelly and the chips just wastes power, AMDs power gating works on the module level not on the integer cores.

On Windows 8, the new cpu scheduler puts the typical workloads on the 1st two modules of the chip, power gates the others, turbo core kicks in more aggresively and the chip runs more cooler. Totally different behavior to Windows 7 and Linux and works very fast too between the cores and modules.

You wouldn't happen to have access to an APU based system to test?. I'm curious if this improved scheduler also applies to the Piledriver based APU's. Perhaps this could get a bit more single threaded performance out of them (i.e. more aggressive turbo)... :hmm:

 

[grimpr]

Unfortunately not, but sure it does work though for the Trinity APUs and more sophisticated since they share power with the igpu.

 

[grimpr]

I'd expect an even greater diff in performance for piledriver(win7 -> win8) that's why I skipped that review, I've read it last year itself.

Definitely, and even more when finding out the compile of lame.exe. I've read reports of linux users running a completely Piledriver compiled gentoo installation using all autovectorization of latest gcc and all extensions (AVX/XOP/FMA3/FMA4/SSE4 etc), they say its fast...thats the beauty of opensource software, living on the true edge instead of x87 and sse2 codepaths in windows programs. Gotta try it someday.:biggrin:

 

[Insert_Nickname]

Unfortunately not, but sure it does work though for the Trinity APUs and more sophisticated since they share power with the igpu.

OK, thanks for the info. 🙂:thumbsup:

 

[Charles Kozierok]

If you've gone to all the trouble of investing in solar panels, which do depreciate and do not have an infinite lifespan themselves, then surely you've gone to all the trouble of computing NPV and NFV costs for your electricity, as well as opportunity costs/profits for selling surpluses back to the utility company.

Seconded. In fact, given that, you have even more reason for going Intel than most people, and it's not like there are particularly compelling reasons to go AMD even for the average enthusiast.

 

[shady28]

Glad you guys mentioned that FX patch. I thought it was installed with normal patches, but apparently not.

I did the email / hotfix thing, got an immediate response and installed on my FX-8320.

Ran before and after PassMark CPU tests, score went from 8391 to 8507. Small 1.4% increase, but every sub-test went up.

Did some more tweaking since I was there, clock-locked the cpu at 4Ghz (stock cooling, so I set the multiplier locked at 20 with no more turbo). Score went up to 9046.

Memory performance was bad for some reason, so I did some timing tweaks and got it stable at 8-8-8-24. Performance there is still just kind of average, Latency was the main killer there (started at 71, now at 56.9, which is still bad). Just from the memory tweaks though, the CPU score went up to 9337.

Max core temp while benchmarking only went up by 2C.

Not too shabby for minimalist tweaking.

 

[tweakboy]

Get it for CCC their great outstanding driver support. Go ATI ... sighs

 

[BuffaloChuck]

If I want more than 2 SATA3 devices in a RAID, AMD's had that solution for 4 years. Intel hasn't delivered on that once.

 

[bononos]

On Windows 8, the new cpu scheduler puts the typical workloads on the 1st two modules of the chip, power gates the others, turbo core kicks in more aggresively and the chip runs more cooler. Totally different behavior to Windows 7 and Linux and works very fast too between the cores and modules.

edited image*
.....

Improvements in windows8 scheduler for BD/PD is more than just squeezing threads into 2 modules and running them at turbo speed. The smarter scheduler will know which threads are better to be separated or lumped together to take advantage of shared L2 cache.

But AMD and/or MS did say very clearly that scheduling improvements will only work for lightly threaded tasks which is why benchmarks are showing only 1-2% improvement in many tests which stress the core. And the i7s with their HT are reaping similar benefit from improved scheduler.

 

[Boze]

So? Power costs. if you feed less to the grid you lose the money you would have gotten from that (assuming that is the case were you live) and due to the fact that solar is a lot more expensive the power consumption difference gets more important.

So I'm not worried about $20 a year more power consumption. I wouldn't be worried about $100 more a year in power consumption.

 

[Charles Kozierok]

If you aren't worried about $100 more per year in power costs, then you certainly can't care about a flat $100 or so savings on the platform. That leaves "because I want to buy AMD" as the only real reason to do so, as far as I can see.

 

[JustMe21]

Unless you overclock and buy high end video cards, power costs aren't that bad among any processor. Let's look at the review of the FX-8350 with power consumption on the x264 video.

http://www.anandtech.com/show/6396/the-vishera-review-amd-fx8350-fx8320-fx6300-and-fx4300-tested/6

Around here, I pay $.11 per kilowatt hour. Now, let's say on average, you use your computer for 5 hours per day during the week and 10 hours per day on the weekend, which would be 45 hours per week or 2340 hours per year.

So, we see the 3570K is using 101.3 W at load, so for 1 year, the total cost would be $26.07 or $2.17 per month.

For the FX-8350, it is using 195.2 W at load, so fo 1 year, the total cost would be $50.24 or $4.19 per month.

Now, obviously you wouldn't be at load the whole time, but even at this load, it's not a $100 difference, but $24.17. Look up an energy cost calculator and see what it looks like based on your power cost in your area.

 

[2is]

Unless you overclock and buy high end video cards, power costs aren't that bad among any processor. Let's look at the review of the FX-8350 with power consumption on the x264 video.

http://www.anandtech.com/show/6396/the-vishera-review-amd-fx8350-fx8320-fx6300-and-fx4300-tested/6

Around here, I pay $.11 per kilowatt hour. Now, let's say on average, you use your computer for 5 hours per day during the week and 10 hours per day on the weekend, which would be 45 hours per week or 2340 hours per year.

So, we see the 3570K is using 101.3 W at load, so for 1 year, the total cost would be $26.07 or $2.17 per month.

For the FX-8350, it is using 195.2 W at load, so fo 1 year, the total cost would be $50.24 or $4.19 per month.

Now, obviously you wouldn't be at load the whole time, but even at this load, it's not a $100 difference, but $24.17. Look up an energy cost calculator and see what it looks like based on your power cost in your area.

Well... That's not really painting a good picture for AMD either seeing that a 3570k is only $20 more and is a better processor for most things. So in a years time you'd have saved $4 over buying AMD and had a better more efficient processor. Not to mention most people are paying more than .11/kwh

 

[Charles Kozierok]

That's what I was saying as well. There's no reason to go AMD over Intel other than saving a few bucks. If even $100 per year doesn't matter, why would saving a smaller amount on a one-time purchase?

 

[Sleepingforest]

Well, maybe it goes into an investment? The value of an upfront lum sum of $100 can be higher than that of $25-50 each year, assuming outrageously good growth.

Disclaimer: I'm not saying this is realistic, but it's sort of a justification maybe.

 

[2is]

Well, maybe it goes into an investment? The value of an upfront lum sum of $100 can be higher than that of $25-50 each year, assuming outrageously good growth.

Disclaimer: I'm not saying this is realistic, but it's sort of a justification maybe.

The cost difference isn't $100. It's $20

 

[Idontcare]

Well, maybe it goes into an investment? The value of an upfront lum sum of $100 can be higher than that of $25-50 each year, assuming outrageously good growth.

Disclaimer: I'm not saying this is realistic, but it's sort of a justification maybe.

If you have an investment opportunity that can make the NFV numbers work out like that then you are a complete fool to even invest in the computer in the first place (Intel or AMD) as you should be putting every nickel you can find into that investment vehicle at that point.

 

[JustMe21]

It's funny how people argue about the power draw of a CPU when a single video card can draw almost 3x the power of the rest of the system. Both Intel and AMD have come down in power usage, although Intel has done a better job. But, I also remember when I had an AM2+ Phenom II X4 940 and an i7 2600 (non K) and I got better response out of the Phenom when working with a lot of large rich text files, PDFs, open office, and picture editor to make e-books.

 

[2is]

It's funny how people argue about the power draw of a CPU when a single video card can draw almost 3x the power of the rest of the system. Both Intel and AMD have come down in power usage, although Intel has done a better job. But, I also remember when I had an AM2+ Phenom II X4 940 and an i7 2600 (non K) and I got better response out of the Phenom when working with a lot of large rich text files, PDFs, open office, and picture editor to make e-books.

You're looking at it out of context. The power draw argument comes into play when people claim they can save so much by going with AMD, when in reality the initial savings is $20 that's all but gone within the first year of ownership.

There are exceptions, like if your primary usage is doing those handful of tasks a processor like the 8350 gives an i7 a run for its money, or if you already have a compatible board and an upgrade is nothing more than a processor swap and may be a BIOS update.

 

[frozentundra123456]

It's funny how people argue about the power draw of a CPU when a single video card can draw almost 3x the power of the rest of the system. Both Intel and AMD have come down in power usage, although Intel has done a better job. But, I also remember when I had an AM2+ Phenom II X4 940 and an i7 2600 (non K) and I got better response out of the Phenom when working with a lot of large rich text files, PDFs, open office, and picture editor to make e-books.

Was the Phenom overclocked?? Very surprised that a quad Phenom at 3.0 ghz could beat a quad Sandy at 3.4 ghz.

 

[2is]

It can't.

 

[MountainKing]

Guys slightly OT, but If an i7 3770k would cost $350 and a FX-8350 $258, which one would you get?

My upgrade would be from a core 2 duo e6600 and mostly for encoding x264 contents from blue ray etc...

 

[frozentundra123456]

Guys slightly OT, but If an i7 3770k would cost $350 and a FX-8350 $258, which one would you get?

My upgrade would be from a core 2 duo e6600 and mostly for encoding x264 contents from blue ray etc...

Personally, I would get a cheaper 3570k because of the low power use and more well rounded performance, at the cost of somewhat lower x264 performance.

If your primary use is x264 encoding though, the FX makes sense. It is slightly faster in second pass x264 encoding than the 3770 and is cheaper. The 3770k is almost as fast in x264, faster sometimes by a large margin in single or lightly threaded benchmarks, and uses less power.

Any of the three would be a huge step up from a core 2 duo however. Depending on how important speed is to you, you might even consider an FX6300. If you are only encoding for your own personal use, you can let the program run overnite, so does a few minutes difference matter to you?

 

[bononos]

.....