• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Return of AMD FX: My OC'd AMD FX 8150 with OC'd 6990 Review - First Results Up!

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Thanks for the feedback everyone! My system becomes unstable if i push nb or ht past 2500ish, and my ram will not jump to 2500 mhz either. It jumps from 2183 mhz to 2475ish, nothing in between for my oc setting. Also i tested my ram with cl 7 settings, but literaly changing it to cl 10 (with nothing else changed) increases my ram performance! This could be an instability issue i am not quite sure.

Tomorrow PC MARK 7 scores! Well see if the patch helps! ;)
 
Yesterday I got new RAM , maybe not really new as someone was using it for a year and I got it for ~45$ with shipping but I think it was worth it. Sticks should be more comparable to yours polyzp - F3-17000CL9D-4GBXMD. So far cpu on water ( not really best set ) and memory on air without fan - http://valid.canardpc.com/show_oc.php?id=2213046.

gskill25.jpg

Don't look on maxxmem results, I can't explain why it's so low :p

I'm starting to think that max memory clock doesn't depends only from IMC quality. This is also PSC set, almost the same IC as in results that I made before on Dominators GT 2000C8 but here is other SPD. I couldn't boot into windows above 2400 on any other PSC sticks unless I was going sub 0 on IMC ( so 10 other PSC sticks ) ... now it's quite stable or at least benchable @2500+, 1.60V and temps like on screenshot ( so for sure not sub 0 ;) ).

I noticed that your 10-12-11 timings are from PSC/4GB profiles from CHV. These aren't best timings to tweak this RAM but at least stable for repetable tests.

Btw, use 2nd and 4th memory slot from cpu side on CHV ( forgot how it was marked ). RAM is more stable and is overclocking a bit better.

I'm not sure why but maxxmem results are weird. I couldn't pass 16000MB/s memory copy with 2500 CPU-NB and 2500 RAM but I could with ~2133 and 2450 CPU-NB ;)
No problems with AIDA64.
 
Last edited:
DIRT 3 benchmarks up!

At first glance it looks pretty impressive but then you see that the FX is clocked @ 4.8Ghz while the 2500K only @ 4Ghz.

It seems that the FX is comparing well against its competition 'again' with the newest games. But it is difficult to gauge because of the way your results have been obtained.

With respect this benching result means little interms of FX vs 2600K.

You would have been better off benching at like for like clock speeds, and to do it with the 2600K as that is around the same cost as the FX-8150.

Form these results it does look like the FX-8150 is at-least comparable to the 2600K clock for clock in Dirt-3 now as well as BF-3.

But we still don't know that, would you be so kind as to take the time to run the benches again clock for clock this time?

Thanks :)
 

Attachments

  • DIRT3-PATCH.png
    DIRT3-PATCH.png
    31.2 KB · Views: 103
Last edited:
TrueCrypt Benchmarks up!

Ill redo DIRT 3 benchmarks by request, and ill test @ 4.0 Ghz. ^^
 
I can't see the point really. The i5 is a fair bit more efficient clock for clock. It would be more relevant to see how high you can clock the i5 using the same cooling solution and test with that. If the i5 can run higher clocks then it is the clear winner.
 
I can't see the point really. The i5 is a fair bit more efficient clock for clock. It would be more relevant to see how high you can clock the i5 using the same cooling solution and test with that. If the i5 can run higher clocks then it is the clear winner.

From my point of view i'm looking at gaming performance, If the FX can match Intel 'there' its not at all a bad thing as the general public are not fussed about clocks as high as 5Ghz, (which the FX-8 will do on watter and the FX-4 on good air anyway)

Generally all people care about is "how much will a BF3, Dirt3..... gamer rig cost me?"

And if there not getting much if any gaming advantages by going 2500K / 2600K then there going to go for the one which costs much less yet is just as good for what they want.

There are enough comments on how well even the FX-4 will run games to counter all the negative press there getting based solely on x86 bench results, enough to make people look at those things with a degree of cynicism, many here will disagree with me on this, but in my mind that's a good thing.

Not everyone is an overclocker enthusiast or interested in x86 bench bragging rights :)
 
Last edited:
It will be rather interesting to see how things improve when AMD has moved across to TSMC's manufacturing process.... Can only be better than Global Foundries..... Likewise, I'm keen on the so called performance enhancements that Piledriver brings forth. It will need to improve both power use at higher clock speeds + have an IPC gain else it will have no chance of keeping up with Intels upcoming offerings.....
 
It will be rather interesting to see how things improve when AMD has moved across to TSMC's manufacturing process.... Can only be better than Global Foundries..... Likewise, I'm keen on the so called performance enhancements that Piledriver brings forth. It will need to improve both power use at higher clock speeds + have an IPC gain else it will have no chance of keeping up with Intels upcoming offerings.....

Indeed, but i thought that was for APU's, i thought PD was still set to be made at GF? Anyway,- who knows how long that will take? TSMC don't have a 32nm bulk node,
they only have 28nm, so as PD was intended for GF it will be 32nm.

It takes some time to DIE shrink and AMD don't exactly have a lot of resources at there disposal,

More on that subject, interestingly TSMC are looking to role out 20nm some time in 2013 <- Bulldozer III ? :D
 
Indeed, but i thought that was for APU's, i thought PD was still set to be made at GF? Anyway,- who knows how long that will take? TSMC don't have a 32nm bulk node,
they only have 28nm, so as PD was intended for GF it will be 32nm.

It takes some time to DIE shrink and AMD don't exactly have a lot of resources at there disposal,

More on that subject, interestingly TSMC are looking to role out 20nm some time in 2013 <- Bulldozer III ? :D

I am guessing Vishera will be the last CPU AMD makes with Global Foundaries. This is Piledriver. With Roughly 15% increase in IPC, and clocks that go as high as 5.4+ Ghz at 10 cores, this will result in an overall performance gain of 40% or so in highly threaded applications. In single core performance, we will only see around 15-20% benefit from current bulldozer offerings which still will not beat Phenom II core Ghz per Ghz, but will wipe the floor with it on highly threaded applications. If a 4.8 Ghz FX core is roughly the performance of a 4.1 Ghz Phenom core, A 5.4 Ghz Vishera core wi atleast perform as good as a 4.4 Ghz Phenom Core. If a Phenom II X6 scales as 5.9, and FX scales as 6.66 , Vishera will scale as roughly 8.3. We can already see that AMD chose not to continue with K10.5 Arch because Bulldozer is headed in the right direction. By the time Vishera is released (Q4/ 2012 ,Q1 2013), most games will take advantage of more than 4 cores.

Ivy bridge on the other hand is said to bring a performance increase of 20-25% overall, with higher overclockability - I am guessing 5.0-5.4 range as with Vishera. Scaling might increase, but even then we can see that even then, intel's mainstream offering will begin to lag behind AMD in highly threaded applications; scaling for the 2600k is 4.87, so mainstream ivy bridge will be around this. I am guessing intel will stay with four cores for mainstream until late in Haswell, but we know that ivy bridge extreme will sport 6 and 8 cores. Intel's current flagship the 3960x scales as 7.23 , and the 2600k scales as 4.87. This implies that an eight core i7 extreme edition ivy bridge will scale as 9.64, which is well above Vishera's scaling of 8.3.

Of course this is all speculation ;)
 
Last edited:
I dont think your ram is scaling at that frequency. also you seem to be throttling or you have all of your power saving features turned on try turning on HPC mode in apm part of AI suite..

for comparison here is mine @ 1866 9-9-9-24 CPu @ 4676mHz
 

Attachments

  • scaling.PNG
    scaling.PNG
    93.1 KB · Views: 75
Symark SiSoftware 2012 Benchmarks up! tomorrow Alien Vs. Predator Revisited!
 
I am guessing Vishera will be the last CPU AMD makes with Global Foundaries. This is Piledriver. With Roughly 15% increase in IPC, and clocks that go as high as 5.4+ Ghz at 10 cores, this will result in an overall performance gain of 40% or so in highly threaded applications. In single core performance, we will only see around 15-20% benefit from current bulldozer offerings which still will not beat Phenom II core Ghz per Ghz, but will wipe the floor with it on highly threaded applications. If a 4.8 Ghz FX core is roughly the performance of a 4.1 Ghz Phenom core, A 5.4 Ghz Vishera core wi atleast perform as good as a 4.4 Ghz Phenom Core. If a Phenom II X6 scales as 5.9, and FX scales as 6.66 , Vishera will scale as roughly 8.3. We can already see that AMD chose not to continue with K10.5 Arch because Bulldozer is headed in the right direction. By the time Vishera is released (Q4/ 2012 ,Q1 2013), most games will take advantage of more than 4 cores.

Ivy bridge on the other hand is said to bring a performance increase of 20-25% overall, with higher overclockability - I am guessing 5.0-5.4 range as with Vishera. Scaling might increase, but even then we can see that even then, intel's mainstream offering will begin to lag behind AMD in highly threaded applications; scaling for the 2600k is 4.87, so mainstream ivy bridge will be around this. I am guessing intel will stay with four cores for mainstream until late in Haswell, but we know that ivy bridge extreme will sport 6 and 8 cores. Intel's current flagship the 3960x scales as 7.23 , and the 2600k scales as 4.87. This implies that an eight core i7 extreme edition ivy bridge will scale as 9.64, which is well above Vishera's scaling of 8.3.

Of course this is all speculation ;)

gloflo are set up and have the experience for CPU's where as TSMC make GPU's, i guess thats why TSMC got the APU contract as its essentially a hot-wired GPU.

Now the thing about the FX-8 and its efficiency is somewhat misunderstood, the thing really does have 8 processor cores in it, (tho not here) some like to argue its like a SB in that it has 4 module and 8 threads, they forget that BD-8 has 2 cores per module, 2x 4 cores = 8 cores, SB has 1 core per module and 4 module, 1x 4 = 4.

Something that has twice the number of cores is going to suck up a lot more power and produce a lot more heat, BD is no where near twice as power hungry as SB nor does it create a massive amount more heat.

All things considered (8 cores) the efficiency of its design is actually quite decent.

If you look at the FX-4100 (which has 4 cores) will clock to 5Ghz on hair without much hassle.

Asking 4 core comparable efficiency from an 8 core is a big ask.

AMD are obsessed with cramming as many cores as possible onto a single DIE, they want 10 cores on there next mainstream CPU, why? how multithreaded do they think APP's are going to get?

I'm one of a group of coders who have spent the last 2+ years coding a new type of software, we are coding for 4 threads which puts it very much in the minority and the reason for that i can tell you is because it is easier just to make do with 1 or 2.

Intel have judged the market much better and stuck to to making each Core more efficient, as opposed to the entire CPU as a whole.

BD may well crush a 2600K if it was using all 8 of its cores simultaneously, the problem is there is nothing which will ask it to do that, having said that with only one cache line for 2 cores that must surely create a bottle-neck eventually.

As for PD's predicted performance improvements? don't hold your hopes up :)

@ salsoul

Here's a Thuban / entry level Corsair RAM memory performance perspective. (3.8Ghz)
 

Attachments

  • mem.PNG
    mem.PNG
    163.5 KB · Views: 78
Last edited:
DIRT 3 revisited ..AGAIN! FX @ 4.0 Ghz. Tomorrow Fritz Chess benchmark, then after Cinebench 10, then SpecviewPerf 11
 
Back