FX-8350 Benchmarks and review

I believe it has something to do with wPrime and SuperPi being based on x87 code (or something like that), which Intel cores traditionally have always done much better in than AMD cores. Which, btw, is also almost no modern programs use.
 
I would not use superpi to compared AMD to Intel. But amd chips to amd chips... 20 seconds doesnt seem any better to me. the other benchmarks seem a tad better.
 
Tsumi said:
I believe it has something to do with wPrime and SuperPi being based on x87 code (or something like that), which Intel cores traditionally have always done much better in than AMD cores. Which, btw, is also almost no modern programs use.
Click to expand...

buttons said:
I would not use superpi to compared AMD to Intel. But amd chips to amd chips... 20 seconds doesnt seem any better to me. the other benchmarks seem a tad better.
Click to expand...

Okay. But that's about the same SuperPi as a stock 8150, and still slower than a 1100T. Cinebench and the others do seem better though, so hopefully that's an aberration.
 
Last edited:
The changes AMD made to Piledriver were mostly on the front end. Reducing branch mispredictions, things like that. If a program does not suffer from a lot of branch predictions, it would benefit a lot less from the improvements AMD made. The typical program does have a lot of misses though. I think the straightforward wprime and superpi do not have a lot of misses.
 
So far its as expected. About a 5-10% increase in single threaded clock for clock over BD, which puts single threaded clock for clock against PHII about equal, maybe a bit behind still. But when OCed all the way, should be trading blows depending on the app.
Multi threaded performance has increased even more.

So productivity users, and heavy multi taskers would gain use from this. Gamers looking for the "best" should stick with their PHIIs, or go to Intel. Not that gaming performance is effected much by CPUs these days with high res, and graphics settings.
 
Last edited:
I'm really surprised they handed these things out when I havent even heard of a review site getting a hold of any
 
Review sites are probably under NDA until release date.
 
I think the performance is mostly due to the way windows 7 handles the cores. Once windows 8 hits I believe the scores will be MUCH higher then they are now
Click to expand...

Windows 7 doesn't really hurt the total speed of Bulldozer when overclocked. Only time it really plays a factor is when you are using stock clocks and make use of the turbo function. Windows 7 has a nasty habit of making threads switch between cores seamlessly and often. The result is for example bulldozer has a turbo core speed of 4.2ghz for single thread applications, but on windows 7 it is switching tasks to other cores which are clocked slower. Thus is not making effective use of the turbo. Second problem Windows 7 has with bulldozer is it will assign tasks to cores, not modules. This plays a role on lightly threaded applications that use less than 4 cores. On the bulldozer arch if you are using less than 4 cores you want to be using one core per module, rather than using a core that shares resources with another core. Of course when you use all 8 cores then it doesn't matter.

The first problem is fixed by overclocking, setting all the cores to the same speed, and not using turbo. Now windows can switch threads to what ever core it wants and they are all the same speed. The second problem was only partly addressed, with the hotfix. The hotfix treats the Bulldozer cpu like a intel cpu. 4 cores 8 threads. Helps some, but still will assign two threads to the same module even though the other modules are not doing anything.

Windows 8 has a different scheduler, it may or may not work properly with the Fx line of cpus. If it does, i wouldn't expect any huge gains, if it doesn't well it cannot be worse the the windows 7 one right? I would run benchmarks on windows 8, but i hate windows 8 so much i refuse to run it.

And yes review sites are still under NDA. I can tell you it won't be much longer.
 
Last edited:
I think, your CPU after OC is trothling or problem is with APM settings...some scores at 4.5 GHz looks strange - winrar worse, superpi worse tahn my fx-8150 at stock....
bios12.jpg


Some boards has not this settings in BIOS then you must try disable it in Windows with AOD.
more info here (from my review)
 
Last edited:
Lord of Shadows said:
I'm really surprised they handed these things out when I havent even heard of a review site getting a hold of any
Click to expand...

Tsumi said:
Review sites are probably under NDA until release date.
Click to expand...

Looks like a new release strategy from AMD, probably caused by having fired most of the marketing department, this CPU has not been hyped as bad as bulldozer and by giving a few enthusiast the chip shortly before launch sets a reasonable amount of expectations at launch.

Review sites are under NDA, but one of the guys from techreport was at the IDF AMD bunker a few weeks ago. He was allowed to say AMD has a lot of stuff close to launch, but he could not mention anything specific.
 
Hey Texanman, can you post the CPU-z "Memory" tab? sPI times rely heavily on memory speed/cycle times. I'll bet that IMC can handle a lot more speed than it's set at currently & tighter tRAS/tRC definitely lower calculation times of sPI. Curious if you can get that FSB @ or passed 300MHz; more interested in +300MHz since Phenom II's can hold 300MHz.

---FSB meaning BUS Speed in this case.
 
Last edited:
Hey Texanman, can you post the CPU-z "Memory" tab? sPI times rely heavily on memory speed/cycle times. I'll bet that IMC can handle a lot more speed than it's set at currently & tighter tRAS/tRC definitely lower calculation times of sPI. Curious if you can get that FSB @ or passed 300MHz; more interested in +300MHz since Phenom II's can hold 300MHz.

---FSB meaning BUS Speed in this case.
Click to expand...

Bulldozers have no trouble with that sort of bus speed. Bus speed does nothing for performance anymore though, unless your also increasing the hyper-transport speed, and northbridge speed as result of increased FSB. FSB is more for fine tuning the last bit of your overclock to achieve more MHZ on the cpu in smaller steps.
 
Lord of Shadows said:
I'm really surprised they handed these things out when I havent even heard of a review site getting a hold of any
Click to expand...

its more of a "create a buzz on the internet" strategy. get people talking about it without having to spend millions of dollars advertising it. but at the same time it could backfire on them but at least they didn't throw a shit ton of money at it like they did when they first released bulldozer. either way AMD saves money, they get free advertising, and a bunch of people got some free processors.
 
Yea 1700/2200 is a tremendous down clock on both values although I can't argue too much w/the "stock" cycle clocks. I'm guessing that DDr frequency is the 18xxMHz choice from the drop-down in the BIOS? Will 9-9-9-27-36 hold stably @1800? 1st priority is to get the NB and "HT" freq well into the 2k's using something around 2400"HT"/2600NB to start. I've seen 2800MHz NB on FX chips and I'll bet it can get to 3200 without too much additional voltage. Pushing a Phenom II @2700MHz NB using only 1.237v CPU/NB and 1.15v NB voltages. Can sPI 1M in the mid 17's although the last few runs have been 18.03xx for some reason. Possibly because I'm not stopping bg apps etc. anymore while testing.

--Added CPU-z can validate if need-be. Just p95'd it like 2days ago and have been gaming pretty heavily since then xD
CPU_z3900_zpsd753e828.png


--Interesting that the TDP says "88" might be because I envoked it 2x before a complete load of the 1st instance. Definitely a 125w CPU tho...weird >.>
 
Last edited:
So in summary
so he got up to 4.5 ghz but not stable..
he has a pretty decent ram speed.

Using an air cooler but not water cooler.

In comparison to my older gen:
I was able to pull my FX4100 to 4.2 ghz with watercooling and ram at the speed of 2133mhz or as CPU Z says frequency of 1070 mhz on the same screen he posted.

Does the vishera have a faster support than 1866 mhz
 
campcreekdude said:
So in summary
so he got up to 4.5 ghz but not stable..
he has a pretty decent ram speed.

Using an air cooler but not water cooler.

In comparison to my older gen:
I was able to pull my FX4100 to 4.2 ghz with watercooling and ram at the speed of 2133mhz or as CPU Z says frequency of 1070 mhz on the same screen he posted.

Does the vishera have a faster support than 1866 mhz
Click to expand...

4.5 is stable. I just never raised the voltage the first time around since i was trying to do soo much with it I just rushed through it and forgot to check other things.
 
Is not problem guys...

My FX-8150 running at 4840 MHz 1.45V stable (yes, is it bit better of piece), 2300 MHz 8-10-8-1T RAM, I have some next FXs-8150/8120 and all can hit 4600-4800+ MHz stable at highend air/decent liquid.
RAM over 2400 MHz stable are not problem too. At benchtable system I got 2500 MHz 10-12-12 RAM stable and 2550 MHz benchbale with only 1.7V at DRAM (tridentX)
 
u must have right motherboard and good cooling and some knowledge how tweaking BD. Of course, with some sh*t as Giga-990x-D3 you cant get over 4xxx MHz much. This board is only lowend with bad mosfets cooling, not much good vrm etc. BD is after OC current hungry, phases are hot. So the best way is Crosshair V Formula, or Sabertooth 990FX, at least UD7 or UD5 or M5A99X Evo, nothing else for really high OC.
After you must know CPU limits, CPU/NB, cooling limits and settings as APM master mode, HPC, LLC, CPU current etc etc.
 
FlanK3rXS said:
u must have right motherboard and good cooling and some knowledge how tweaking BD. Of course, with some sh*t as Giga-990x-D3 you cant get over 4xxx MHz much. This board is only lowend with bad mosfets cooling, not much good vrm etc. BD is after OC current hungry, phases are hot. So the best way is Crosshair V Formula, or Sabertooth 990FX, at least UD7 or UD5 or M5A99X Evo, nothing else for really high OC.
After you must know CPU limits, CPU/NB, cooling limits and settings as APM master mode, HPC, LLC, CPU current etc etc.
Click to expand...

I was under the impression the only difference between the gigabyte UD3 and 5/7 were options for multiple graphics cards?
 
I feel like I am going to need to get my hands on one to get some REAL OC'ing results.

Most people (and review sites) are scared to raise vcore much beyond stock, also most review sites are going to be using high end air coolers/low end water coolers. Even if they do use decent cooling they will still probably leave the Northbridge and Hypertransport at stock speeds.

Basically nobody is going to give you a solid review of this CPU's OC ability.
 
leeleatherwood said:
I feel like I am going to need to get my hands on one to get some REAL OC'ing results.

Most people (and review sites) are scared to raise vcore much beyond stock, also most review sites are going to be using high end air coolers/low end water coolers. Even if they do use decent cooling they will still probably leave the Northbridge and Hypertransport at stock speeds.

Basically nobody is going to give you a solid review of this CPU's OC ability.
Click to expand...

This is true. For me I just do not want to go into the high risk of something failing. I have a limited budget being a student and can not afford replacements if something were to break on me.
 
I was under the impression the only difference between the gigabyte UD3 and 5/7 were options for multiple graphics cards?
Click to expand...

yes that is correct.

Ud3 is every bit as capable as the ud5, and ud7. Though it does have a tad bit weaker VRM/northbridge cooling.

I feel like I am going to need to get my hands on one to get some REAL OC'ing results.

Most people (and review sites) are scared to raise vcore much beyond stock, also most review sites are going to be using high end air coolers/low end water coolers. Even if they do use decent cooling they will still probably leave the Northbridge and Hypertransport at stock speeds.

Basically nobody is going to give you a solid review of this CPU's OC ability.
Click to expand...

I will push to 1.55volts, no higher though. Most review sites will push to 1.45 or maybe 1.5. Overclocking the northbridge and hypertransport doesn't really do much in the way of performance on Bulldozer like it did with Thuban or Phenom 2.
 
Texanman said:
This is true. For me I just do not want to go into the high risk of something failing. I have a limited budget being a student and can not afford replacements if something were to break on me.
Click to expand...

Thats completely understandable. Review sites though get them for free and have lots of other systems for backup, so they should push them.
 
Nice article - I never seem to be that lucky. I was just wondering what kind of BIOS modifcations you had to make and what was you BIOS version?

Thanks,

Tom D
 
ZZZTOPZZZ said:
Nice article - I never seem to be that lucky. I was just wondering what kind of BIOS modifcations you had to make and what was you BIOS version?

Thanks,

Tom D
Click to expand...


No bios mods needed. I am running the latest version (1.80) on my asrock board
 
buttons said:
I would not use superpi to compared AMD to Intel. But amd chips to amd chips... 20 seconds doesnt seem any better to me. the other benchmarks seem a tad better.
Click to expand...
A processor's job is to compute math.SuperPi is a math calculating program.
 
Technically, that statement is horribly incorrect. Everything is 'in the real world'. Perhaps you mean 'daily use'? For instance, "for people that use M$ office every day this doesn't show how negligible Intel's performance lead may be", since if I wrote something that was optimized for something particular to Intel chips and thus it performed poorly on AMD chips that would still be a 'real world' scenario.

I still don't buy the argument, since Intel SB/IB simply walks all over AMD when it comes to single-threaded performance with pretty much any application. At least they're kicking ass in multi-threaded things a little bit harder now.
 
SuperPi uses the x87 code path, which Intel has always been better than AMD in. Also, virtually nothing uses x87 nowadays.
 
Hi, thanks for the early benchmarks man :)

but i would like to see the following benchmarks please !!! im begging you!!!

3dmark11 benchmark = want to know how much a difference we get with the FX8350

Power Consumption = compared to FX8150 did they fix the power consumption?

thanks
 
The gig ud3 is very different then the ud5/7. It's not just the cooling. It has the same older vrm and power delivery that the 7 and 8 series boards had. It's a great board for phii, but it can't even handle an fx-8120 at stock if you disable the throttling. They don't have an option for it in the bios because it can't handle it. I did a work around to do it, and it killed both the board and CPU. Again at stock speeds, using prime95. Vrms and nb were pushing well past 100c.
 
EliteRepLaY said:
Hi, thanks for the early benchmarks man :)

but i would like to see the following benchmarks please !!! im begging you!!!

3dmark11 benchmark = want to know how much a difference we get with the FX8350

Power Consumption = compared to FX8150 did they fix the power consumption?

thanks
Click to expand...

Ill update my gaming power usage tonight. I always forget to look down at the kill-a-watt in the heat of battle :D I have my idle temps in the first post. (for comparison the same setup i have Just with my older 965 the system was ~135-145w idle but I was not overclocked. that 160w i have in the first post is me overclocked to 4.5ghz no turbo)
 
You must log in or register to reply here.
Back
Top