Video benchmark of FX 8150(not good)

What a shitty video. They never even show the actual system being tested or the CPU, just a photoshopped one.
 
Three reviews (plus hands on accounts) showing the same thing. If this "modern" processor performs like this, how in the hell is AMD hoping to get $200+ for this thing?

Are they praying folks get suckered into purchasing with the "real 8 cores" claim? Even their previous generous models are giving BD a run for its money.

Pure comedy if this turns out to be the story on Wed.
 
fiat-cinque-tank.jpg
 
lol that thing in the pic could plow my drive this winter. I got $200
 
spydermonkey said:
I don't see any bulldozer CPUs listed in the supported CPU section. Hopefully there's some explanation as to why these figures suck. I really hope so, anyway. I'm going to remain cautiously optimistic until Wednesday.
Click to expand...

Well, BD hasn't been released yet, so none of the board manufacturers (that I know of ) show their boards supporting BD in the cpu support lists yet. However, Agesa has already been released.

BTW, don't mean to troll, but had to post this:

http://www.youtube.com/watch?v=CqTU4wVvZL0&feature=youtu.be
 
with all these user reviews out already, I hope AMD moves up the embargo date to Monday.

every member of the press who has this chip probably has the review all typed up and ready to go already.
 
Why is everyone so pissed with the 8150 running within 1% of the 2600k? I didn't think anyone was still expecting it to beat the 2600k at stock.
 
lasserith said:
Why is everyone so pissed with the 8150 running within 1% of the 2600k? I didn't think anyone was still expecting it to beat the 2600k at stock.
Click to expand...

they are pissed because it is essentially a phenom x8.
AMD could have released a 150w phenom x12 two years ago, and it would have stomped everything Intel in multithread tasks.

I'm sure everyone was expecting a big increase in IPC, but in the end, AMD just threw more cores at the task. In weakly threaded tasks, AMD's shortcoming is going to stick out.
It took a 3.6ghz amd to keep up with a 2.6ghz first gen core i7 in single thread tasks. Given the ipc improvement in SB, it would take a 5ghz FX to keep up with a stock 2600k in per core performance.
Once you OC that 2600k, FX will just eat dirt.

The FX is essentially best for people who exclusively use infinitely threaded apps.
Until most apps can scale to infinite cores, Intel is still king.
 
According to this benchmark, it trades blows with 2600k. It beats it in at least 3-4 non synthetic benchmarks. I don't see what the problem is.
 
sirmo said:
According to this benchmark, it trades blows with 2600k. It beats it in at least 3-4 non synthetic benchmarks. I don't see what the problem is.
Click to expand...

According to DonanimHaber, pigs can fly and computers run on Swiss cheese.
 
heflys20 said:
If I were Intel, I'd raise the price of SB.
Click to expand...

Intel to the world : " Considering that AMD charges you almost 2600k price for a lame duck, we believe you're getting a too good deal for the 2600k. New price effective tomorrow is $399."
 
lasserith said:
Why is everyone so pissed with the 8150 running within 1% of the 2600k? I didn't think anyone was still expecting it to beat the 2600k at stock.
Click to expand...

It's not. The difference is 20-40%.

The 1% comes from GPU limited scenarios, playing at 1080P or 2560x1600. If I put together a man and an ellephant to see who draws more air through a straw in 5s, what does that tell you about their respective strengths ? Nothing really.
 
savantu said:
It's not. The difference is 20-40%.

The 1% comes from GPU limited scenarios, playing at 1080P or 2560x1600. If I put together a man and an ellephant to see who draws more air through a straw in 5s, what does that tell you about their respective strengths ? Nothing really.
Click to expand...

This was THE worst analogy of, probably, all-time.
 
savantu said:
If I put together a man and an ellephant to see who draws more air through a straw in 5s, what does that tell you about their respective strengths ? Nothing really.
Click to expand...

What does that have to do with the price of tea in China?
 
Just wait for official release and official benchmarks. Only a couple more days. But to be honest I don't think Bulldozer will be great because there are a few reviews already saying it isn't too great.
 
Question: If the Bulldozer was only 4 cores and had the same performance, would your opinion change?

I know AMD is trying to spin their design as 8 cores but they ain't fooling anybody; its quad core hyperthreading on steroids but their patent laywers say it ain't.
 
old skool said:
Question: If the Bulldozer was only 4 cores and had the same performance, would your opinion change?

I know AMD is trying to spin their design as 8 cores but they ain't fooling anybody; its quad core hyperthreading on steroids but their patent laywers say it ain't.
Click to expand...

Cause it ain't. The shared FPU is shared only in 256-bit AVX mode. In 128-bit SSE mode (which is what all software is now) it splits into two units, thus each module is one integer and one floating point unit.
 
iconiK said:
Cause it ain't. The shared FPU is shared only in 256-bit AVX mode. In 128-bit SSE mode (which is what all software is now) it splits into two units, thus each module is one integer and one floating point unit.
Click to expand...

And if these "cruddy leaked benchs are 4 realz" then that design leads to almost a 10 percent deficiency in IPC(PC) per core over Phenom 2 BE CPU's which might mean it has more trouble (a 10 percent deficiency over Phenom architecture IPCPC) in the fetch and decoding stages or somewhere else in the architecture.

Because the 8 core 8120 was tested @ 3.1Ghz it stated a 25second Pi results, and my X3 @ 3.46GHz (yet lacks Level 3 cache) I get exactly 24.497 Pi). 10 percent less IPC from 3.46Ghz 346mhz. So 3.1+ 346mhz = 3,446 GHz. If the average cache on a Phenom 2 gets 5 percent then it's 15 percent slower than my chip so I can get 4Ghz on my chip I would need Bulldozer to OC 15 percent more than Bulldozer to see the same IPCPC (if bench's are real) 4.0GHz x .15 = .600. SO for me to get same IPCPC out of each Bulldozer COre I need an Overclock of 4.6 GHz 24/7 to just match IPCPC performance of my X3 @ 4.0Ghz. but i will also get 4/5 more cores

If this chip does 5.2GHz easy for me it's an IPCPC and 2x Core increase all around. IPC @ 5.2 GHz is 10 percent better IPC per core than my current X3 that unlocks to X4, But that bump begins dumping 20 degrees more heat which I hate in the summer so I usually run it @ X3 to avoid the additional heat.

Add also if it (BD 8 Cores) can get to 5.8GHz heck that's 20 percent IPCPC over this X3/X4 that would be amazing IMO for 250 buckaroos.

Can't wait to see this thing REALLY benched! It is a solid upgrade for me if this is true and it OC's like a champ....at least I think :p I really would like 2 know the max stable OC on these FX 8 cores on air cooler like the Noctua NH-D14.
 
Last edited:
iconiK said:
Cause it ain't. The shared FPU is shared only in 256-bit AVX mode. In 128-bit SSE mode (which is what all software is now) it splits into two units, thus each module is one integer and one floating point unit.
Click to expand...


No, that is not true. A single-threaded application can issue 2x SSE instructions per clock. If you have 2 multimedia threads running on the same module, they will share.

It would be more accurate to say it this way: Bulldozers integer units are true cores, while the floating point units are hyper-threaded. Call it "Hybrid-Threading" if you want, but AMD's patent lawyers insist that you don't call it hyper-threading.
 
teletran8 said:
And if these "cruddy leaked benchs are 4 realz" then that design leads to almost a 10 percent deficiency in IPC(PC) per core over Phenom 2 BE CPU's which might mean it has more trouble (a 10 percent deficiency over Phenom architecture IPCPC) in the fetch and decoding stages or somewhere else in the architecture.
Click to expand...


The IPC deficiency can be more extreme than you think; AMD K10 can execute 3x identical ALU (that means integer) instructions per clock; although Bulldozer can execute up to 4x, it must be a mixture of loads, stores and arithmetic. Where you had 3 IPC on the K10 you may wind up with only 2x on the Bulldozer. In the real world, however, it lands somewhere in the middle (and maybe ahead) due to the nature of x86 code; it just so happens the small register set of x86 makes it necessary to load and store intermediate values often.


AMD isn't trying to smash performance barriers with Bulldozer. They are re-thinking the microarchiture to achieve maximum performance-per-watt and I think they achieved that.
 
old skool said:
The IPC deficiency can be more extreme than you think; AMD K10 can execute 3x identical ALU (that means integer) instructions per clock; although Bulldozer can execute up to 4x, it must be a mixture of loads, stores and arithmetic. Where you had 3 IPC on the K10 you may wind up with only 2x on the Bulldozer. In the real world, however, it lands somewhere in the middle (and maybe ahead) due to the nature of x86 code; it just so happens the small register set of x86 makes it necessary to load and store intermediate values often.

AMD isn't trying to smash performance barriers with Bulldozer. They are re-thinking the microarchiture to achieve maximum performance-per-watt and I think they achieved that.
Click to expand...


From these leaked benchs or whatever it does seem that is the "downside" to this architecture in that it shares alot of resources and does most everything well yet there is a "Little impact on timing and complexity of critical paths" as stated in August 2010 slide from AMD.

bulldozer-concept.jpg


Yet the upside to that coin is unlike every i5/ some lower end i7. AMD can squeeze double the cores (8 AMD vs 4 Intel) on a similar CPU die size in the same "ballpark CPU price range" compared to Intels i5/some i7's. Yet Intels IPCPC is what like 50+ pct faster only in Hyper Pi I think, maybe in 3D mark 2001 too I dunno? 10 sec 1M runs SB 2500k and up CPU's vs BD 8120 @ 3.1GHz of 25 second Hyper Pi runs. The FX 8150@ 3.6GHz would be closer to 22 I would think maybe even close to 20sec. I dunno we don't have the data yet so it's guessing, but yeah Intel has AMD beat on IPCPC but sells just half the cores in that CPU price range.

These companies are taking 2 different directions really if you think about it that way.

IMO I would say BD takes a 10 percent hit between PII and BD1 architecture from pre-bench result tests in Hyper Pi. Gaming at high/ultra settings seems unaffected if not a performance increase over PII in min and max fps, which is a good thing for PC gamers.

X6 1090t and 1100t owners chips come @ 3.2GHz and 3.3GHz but the 8150 is supposed to be stock @ 3.6 which basically overcomes that 10 percent deficit to match up core for core. So a non-OC'd X6 user with those CPU's would only see the extra performance of having 2 more cores, since the 10 pct performance loss is overcome/balanced by the 3.6 stock clock speed of an FX 8150. Hopefull FX 8150 OC's more than 1 GHz over stock on good cooling or else I don't see why X6 users should upgrade considering very few games do use 6 cores today....yet there's always tommorrow...hmmmm
 
Last edited:
teletran8 said:
And if these "cruddy leaked benchs are 4 realz"...

Because the 8 core 8120 was tested @ 3.1Ghz it stated a 25second Pi results
Click to expand...

Something doesn't seem right... I mean this Q is an older chip, but the IPC....

SuperPI isn't a real good indicator on a 8 core chip anyway. I have a Q9550 @2.83 I'll get 16.599 s on 1 mil....

Cinebench - 3.45

wPrime - 32m 15,756s. 1024m 497,218

Fritz 4.2 - 16.5 Relative speed. Kilo Nodes p/sec 7919

x264 Benchmark _HD v4.0. Highest 1st pass 85.67, 2nd pass 19.46.

Multicore benches of course it gets destroyed but...

Anyone have an i7 that they can run these benches on?
 
lasserith said:
Why is everyone so pissed with the 8150 running within 1% of the 2600k? I didn't think anyone was still expecting it to beat the 2600k at stock.
Click to expand...

They're not really pissed more disappointed since the wait was filled with so many delays and the reward was running within 1% of its competitor which has been out for 8? months. They were expecting an 8 core monster that would destroy the competition what they got was meh
 
old skool said:
AMD isn't trying to smash performance barriers with Bulldozer. They are re-thinking the microarchiture to achieve maximum performance-per-watt and I think they achieved that.
Click to expand...

Well they're still in 125W TDP land with high end models, and with mediocre yields, that will be close to what you'll see most of the time. Intel has 80W (15W is the GPU) but the good yields means the CPUs don't get too close to the TDP. Though Intel has a much larger R&D budget, so it's understandable.


prometheum said:
They're not really pissed more disappointed since the wait was filled with so many delays and the reward was running within 1% of its competitor which has been out for 8? months. They were expecting an 8 core monster that would destroy the competition what they got was meh
Click to expand...

Except it runs far slower than 1%. The 1% results are GPU-limited.
 
You must log in or register to reply here.
Back
Top