Saturday, September 24th 2011

AMD FX 8150 Looks Core i7-980X and Core i7 2600K in the Eye: AMD Benchmarks

The bets are off, it looks like Intel is in for a price-performance shock with AMD's Bulldozer, after all. In the press deck of AMD FX Processor series leaked by DonanimHaber ahead of its launch, AMD claims huge performance leads over Intel. To sum it up, AMD claims that its AMD FX 8150 processor is looking Intel's Core i7-980X in the eye in game tests, even edging past it in some DirectX 11 titles.

It is performing on par with the Core i7-2600K in several popular CPU benchmarks such as WinRAR 4, X.264 pass 2, Handbrake, 7Zip, POV Ray 3.7, ABBYY OCR, wPrime 32M, and Bibble 5.0. AMD FX 8150 is claimed to be genuinely benefiting from the FMA4 instruction set that Sandy Bridge lacks, in the OCL Performance Mandelbrot test, the FX 8150 outperforms the i7-2600K by as much as 70%. Lastly, the pricing of the FX 8150 is confirmed to be around the $250 mark. Given this, and the fact that the Core i7-2600K is priced about $70 higher, Intel is in for a price-performance shock.


Source: DonanimHaber
Add your own comment

854 Comments on AMD FX 8150 Looks Core i7-980X and Core i7 2600K in the Eye: AMD Benchmarks

#1
mastrdrver
Two things I would be interested in seeing: performance in Crossfire/SLI and with a higher clocked NB.

Phenom IIs are not very good at scaling performance when you have two (or more) GPUs. I also wonder how much/high you can overclock the NB part. If I understand correctly, all of SB cores and caches have a single clock. If that is true then I wonder how much overclocking the BD NB would help its performance?
Posted on Reply
#2
crazyeyesreaper
Chief Broken Rig
going from 2200nb to 2400nb gives 1000mb/s memory bandwidth improvement

so say Bulldozer at 1866mhz gets 18gb/s memory bandwidth a 2400mhz NB will bump it up to 19gb/s

so right now 200mhz NB clock speed bump gives 1000mb/s memory bandwidth bump

problem is on an overclocked 8150 at 4.6-4.7ghz the NB tops out at 2400mhz meaning

the NB was only capable of a 200mhz overclock while the CPU cores were cable of a 1ghz overclock
Posted on Reply
#3
Inceptor
crazyeyesreaper said:
problem is on an overclocked 8150 at 4.6-4.7ghz the NB tops out at 2400mhz meaning
the NB was only capable of a 200mhz overclock while the CPU cores were cable of a 1ghz overclock
At what bus speed?
At what NB multiplier?
At what NB voltage?

You know, I haven't made the decision whether or not to buy BD yet, but uhh... who cares about a 1 ghz cpu overclock. I don't care what the UEFI/BIOS auto settings set the NB frequency to, I want to see what is possible with a manual increase. A cpu multiplier overclock is the simplest possible way to do it. I want to see someone bench one of these cpus with an increased bus speed, to up the NB frequency.

I thought you said you were done?
Posted on Reply
#4
crazyeyesreaper
Chief Broken Rig
its Laughingmans review for Bulldozer over on futurelooks its not up yet but im in TS3 with him

2200nb is Bulldozers default

2400nb aka 12x200 = 2400nb

that was all it would do

litterally set the NB any higher board wont post wont boot nothing blah done.


nb clocking isnt the BD chips forte cpu core clocks are.

and im not ALWAYS trolling alot of the info ive posted in this thread is actually truth, almost all of it is , i just played it to a very optimistic tune to piss ppl off

that said again

2400nb was all he could get out of the chip up from 2200 but it did give a 1000mb/s bandwidth increase at 1866mhz ram so it went from 18gb/s to 19gb/s its an improvement still not as good as Sandybridge tho

overclocked sandy at 1866 puts out 24-25gb/s but still far better then Phenom II's memory bandwidth but it hardly effects any of the benchmarks it did make a difference in cinebench but not a large one not enough to beat a 2600k anyway not anywhere near close

basically in 99% of tests nb speed didnt effect performance at all, it gave a nice bump to memory bandwidth but it didnt translate to better performance.
Posted on Reply
#5
Inceptor
That's @200mhz.
I'm talking about higher bus speeds and higher NB voltage.
Did he try upping the NB voltage?
Posted on Reply
#6
crazyeyesreaper
Chief Broken Rig
yes he did it didnt post whats so hard to understand about that

from what ive seen NO one has touched the northbridge period except for Laughingman lol im guessing thats the reason for it.

basically

your stuck at interals of 2 on the NB

2200
2400
2600
2800

at max overclock on the cpu the nb wont go above 2400 system dosent boot, not clock speed adjustments might make more of a difference but from what im looking at its not bringing any real performance gains, altho i can say Laughingmans review will probably be the only one around with NB clocks in the mix

I found what you wanted

http://www.madshrimps.be/articles/article/1000220/AMD-FX-8150-Bulldozer-CPU-Review/6#axzz1aYGbhypH

2200 2600 3200 NB speeds

guess what performance benefit = 0%

memory bandwidth increased but thats it no benefit otherwise.
Posted on Reply
#8
ivicagmc
That they just shrink Phenom to 32nm, with few tweeks, packed in 8 real cores it would be much better, and earlier. WTF AMD engineers were doing? This is disaster and fail not compering to SB, but compering to Thuban... And I have spent my hard earned money and bought AM3+ board. AMD you have fail me for the last time. As soon as I get some money I'm going to the dark side...
Posted on Reply
#9
Mussels
Moderprator
i want to see some DX11 benchies, as the multi threadedness there should really help.

DX11 in crossfire/SLI would be best.
Posted on Reply
#10
de.das.dude
Pro Indian Modder
as always AMDs problem lies with its memory controller. some one isnt working hard enough.
Posted on Reply
#11
HalfAHertz


AMD needs lots of it^

Well this blows. I was looking forward to adding a new AMD PC to my makeshift "rendering farm". It will still be an AMD CPU but from the looks of it I'll stay clear from Bulldozer.
Posted on Reply
#12
TRWOV
So the best buy is going to be the FX-8120. Being the same silicon it should overclock almost as good as the FX-8150.

Next steppings and kernel patches just can't come fast enough. I think that this platform will need some revisiting by year's end to see how much performance is to be expected with appropriate software.


The good:
- Decent multi-thread performance depending on the application.

The bad:
- Lower clock for clock performance than Thuban.
- Requires software patches.
- Needs overclocking out of the box to consistently beat the 2500K and 1100T at stock clocks.

The ugly:
- The power bill.



Grab a PII x6 while you can if you don't have it already. Later on, if the platform evolves to something worthy you can upgrade then.
Posted on Reply
#13
HalfAHertz
Well the few good things about it is that it has Hardware AES-Ni support, TurboCore that actually works and low power consumption(for its gynormous size) at idle thanks to power gating at the silicon. So it's finally on par with $ntel in those fields and further behind in the rest :(
Posted on Reply
#14
btarunr
Editor & Senior Moderator
Much of the ~2 billion transistor count is accounted for by the 16 MB cache.
Posted on Reply
#15
cadaveca
My name is Dave
btarunr said:
Much of the ~2 billion transistor count is accounted for by the 16 MB cache.
Yeah, which also explains the high power draw. Makes me wonder jsut how good per-core perforamcne is with a CPU core that small.

Really, liek I said a couple fo days ago, the large amount of cache kinda hinted that bulldozer jsut couldn't live up to the hype some had placed on it. Everything is quite obvious, and has clear reasons why things are the way they are.


I jsut don't get why everyone is so surprised. Maybe next time when I say to ignore the hype, more people will listen.
Posted on Reply
#16
AhokZYashA
so, BD is marginally slower than SB in most application?
Posted on Reply
#17
Shihabyooo
AhokZYashA said:
so, BD is marginally slower than SB in most application?
To put it bluntly, yes.

*Sigh* And with the SB-E unpromising reviews, will any 2011 CPU release worth the bother ?
Posted on Reply
#18
HalfAHertz
AhokZYashA said:
so, BD is marginally slower than SB in most application?


On average 8,8% slower than 2500k(not counting superPi coz it's ghey) and 19% slower than 2600k according to hardwarecanucks

What's more appalling is how much "faster" it is than it's predecessors:



Again on average and not counting superPi: 11,9% faster than a PhII 980 and the whopping 2,6% faster than a 1100T

The moral here, at least ofr me, is that the 1100T was already competing quite nicely with the 2500k in multi-threaded workloads and what AMD needed to mainly focus on was their IPC and IMC.

Edit: sorry for my thread crapping in all the BD threads but I'm simply lost for words here :(
Posted on Reply
#19
TheMailMan78
Big Member
cadaveca said:
Yeah, which also explains the high power draw. Makes me wonder jsut how good per-core perforamcne is with a CPU core that small.

Really, liek I said a couple fo days ago, the large amount of cache kinda hinted that bulldozer jsut couldn't live up to the hype some had placed on it. Everything is quite obvious, and has clear reasons why things are the way they are.


I jsut don't get why everyone is so surprised. Maybe next time when I say to ignore the hype, more people will listen.
I listened. Still not gonna scrap my whole rig for a SB however! YOU CANT MAKE ME! lol
Posted on Reply
#20
Bucknuts77
http://www.guru3d.com/article/amd-fx-8150-processor-review/

The 8150 is a joke, and not even a good one. Sure its a good CPU for a work PC if the price was a bit lower, but this CPU was market as a high end gameing CPU, and is no better than my 1090T @ 4.2, and you OC this BD to 4.6 and your useing over 430 wats. This almost makes me want to rip out my Sabertooth and break it in half.
Posted on Reply
#21
digibucc
how does it compare to pre-sb intels? i7 920/940/965?
Posted on Reply
#22
15th Warlock
digibucc said:
how does it compare to pre-sb intels? i7 920/940/965?
TBH, who cares, that comparison would have been relevant two years ago, when BD was supposed to be released. The present is SB, and SB-E is just around the corner, X58 is EOL as far as upgrade options go...
Posted on Reply
#23
Shihabyooo
digibucc said:
how does it compare to pre-sb intels? i7 920/940/965?
According to tom's hardware's review, the 920 beats BD in many single threaded progs, In gaming, the 920 smothers it. BD managed to outperform the 920 in most multi-threaded apps.
Posted on Reply
#24
nt300
Apparently AMD is complaining about an issue with how Windows Operating systems are reading Bullodzer. It can't seem to figure out the design in terms of its modules and therefore Bulldozer should in reality be approx: 15% faster if Microsoft can figure this out and come out with a patch or something to ensure Bullodzer is fully understood. I believe Guru3D has a much better explanation in this matter.

Microsoft did state that Windows 8 will be fully optimised to take advantage of Bulldozer. In the meantime, do I have an incentive to upgrade my setup with a Bulldozer CPU? Dam this sucks…… :confused:
Posted on Reply
#25
AhokZYashA
that makes my word many pages back true then.

tbh im looking forward to BD performance,
turns out its just a joke, even compared to an older i7 9xx
Posted on Reply
Add your own comment