Fudzilla: Bulldozer performance figures are in

[Davidh373]

I didn't realize that. Thanks for the info.

He's right. I wouldn't get an AM3+ though. Wait until the full featured motherboards come out so you don't miss out on any nifty goodies!

 

[SlowSpyder]

Charlie says it should ship tomorrow.

http://semiaccurate.com/2011/08/04/bulldozer-to-ship-soon-really-soon/

Will we see benches? I wonder how six core chips will be priced.

 

[grimpr]

Most propably he means Opterons and not desktop FXs.

 

[SlowSpyder]

Most propably he means Opterons and not desktop FXs.

I know, but then we should see something if he's correct, right?

We could get some idea of what to expect from the architecture.

 

[inf64]

Links please

There you go :
http://www.chiphell.com/thread-250461-1-1.html
These are the BD ES results. BD ES ran at 3.6Ghz(just as 2600K which was OCed to that speed),no idea if Turbo was on-even worse for the ES results if it was on.
wprime- 10.414s
c11.5- 4.11pts
super pi- 23.91s
fritz chess- 9953pts

Now on to Thuban and Deneb scores. Used for comparison are 3.3Ghz X6(with 3.7Ghz Turbo) and 3.5Ghz/3.6Ghz X4(Deneb). Click on the benchmark name for sources of the reviews.
wprime - X4 975 @ 3.6Ghz 10.8s ; X6 @ 3.3Ghz 8.32s
c11.5 - X4 975 @ 3.6Ghz 4.28pts ; X6 @ 3.3Ghz 5.91pts
super pi - X4 975 @ 3.6Ghz 19.37s ; X6 @ 3.3Ghz 19s
fritz chess - X4 975 @ 3.6Ghz 8825pts X6 @ 3.3Ghz 11677pts

So now that I've provided the links of both X4 and X6 in these computation benchmarks,let's see how much slower is this poor ES compared to age old X4 and a bit younger X6. All chips at their rated clocks : BD 8C @ 3.6Ghz, X4 @ 3.6Ghz, X6 @ 3.3/3.7GHz.

wprime - BD 8C is 3% faster than age old X4 and is 25% slower than X6(AMD targets 300$ price range for this model while it is supposedly 25% slower than 189$ X6? lol). Basically it is the same as quad core deneb in this test. First sign of failure ES results(be it microcode,bios whatever).
c11.5 - BD 8C is 4% slower than age old X4 and is 44% slower than X6 . Another even clearer sign of failed test platform(ES/microcode/insert whatever).
super pi - BD 8C is 23% slower than age old X4 and is 26% slower than X6. More of the same as above.
fritz chess - BD 8C is 12% faster than age old X4 and is 17.3% slower than X6. Rinse and repeat.

Conclusion: the top 8 core FX part(8150 @ 3.6Ghz,with Turbo to 4.2Ghz) that is going to cost approximately 300$ is performing slower or on par with 159$ X4 975 Deneb model(3.6GHz) and is much slower- 30% on average - than X6 1100T Thuban chip that now costs 189$. Anybody else sees the problem here or do I have to draw a picture? :hmm:

These results are bullcrap.Period.

 

[guskline]

Not as silly as you look for talking about TDP figures like that. Also keep in mind that TDP between the two is not directly comparable since each company figures it differently.

Darkswordsman17. Do you own an I5 2500k or an I7 2600K? I own and USE an I5 2500k oc to a safe 4400. It can boot into Windows at 4800 and run most software at 4600. Unless AMD scales the Bulldozer better than the Phenom IIs I wonder how well they will OC. BTW I am well aware that Intel and AMD use different systems for voltage. I also own 2 AMD 965 BE systems OC to 3700. My point is that unless AMD really kicks it up a notch they are going to have big trouble with their 8 core

 

[Avalon]

There you go :
http://www.chiphell.com/thread-250461-1-1.html
These are the BD ES results. BD ES ran at 3.6Ghz(just as 2600K which was OCed to that speed),no idea if Turbo was on-even worse for the ES results if it was on.
wprime- 10.414s
c11.5- 4.11pts
super pi- 23.91s
fritz chess- 9953pts

Now on to Thuban and Deneb scores. Used for comparison are 3.3Ghz X6(with 3.7Ghz Turbo) and 3.5Ghz/3.6Ghz X4(Deneb). Click on the benchmark name for sources of the reviews.
wprime - X4 975 @ 3.6Ghz 10.8s ; X6 @ 3.3Ghz 8.32s
c11.5 - X4 975 @ 3.6Ghz 4.28pts ; X6 @ 3.3Ghz 5.91pts
super pi - X4 975 @ 3.6Ghz 19.37s ; X6 @ 3.3Ghz 19s
fritz chess - X4 975 @ 3.6Ghz 8825pts X6 @ 3.3Ghz 11677pts

So now that I've provided the links of both X4 and X6 in these computation benchmarks,let's see how much slower is this poor ES compared to age old X4 and a bit younger X6. All chips at their rated clocks : BD 8C @ 3.6Ghz, X4 @ 3.6Ghz, X6 @ 3.3/3.7GHz.

wprime - BD 8C is 3% faster than age old X4 and is 25% slower than X6(AMD targets 300$ price range for this model while it is supposedly 25% slower than 189$ X6? lol). Basically it is the same as quad core deneb in this test. First sign of failure ES results(be it microcode,bios whatever).
c11.5 - BD 8C is 4% slower than age old X4 and is 44% slower than X6 . Another even clearer sign of failed test platform(ES/microcode/insert whatever).
super pi - BD 8C is 23% slower than age old X4 and is 26% slower than X6. More of the same as above.
fritz chess - BD 8C is 12% faster than age old X4 and is 17.3% slower than X6. Rinse and repeat.

Conclusion: the top 8 core FX part(8150 @ 3.6Ghz,with Turbo to 4.2Ghz) that is going to cost approximately 300$ is performing slower or on par with 159$ X4 975 Deneb model(3.6GHz) and is much slower- 30% on average - than X6 1100T Thuban chip that now costs 189$. Anybody else sees the problem here or do I have to draw a picture? :hmm:

These results are bullcrap.Period.

Obviously fake results. They're not going to ship a slower product.

 

[RobDickinson]

There not shipping it...

 

[darkswordsman17]

Darkswordsman17. Do you own an I5 2500k or an I7 2600K? I own and USE an I5 2500k oc to a safe 4400. It can boot into Windows at 4800 and run most software at 4600. Unless AMD scales the Bulldozer better than the Phenom IIs I wonder how well they will OC. BTW I am well aware that Intel and AMD use different systems for voltage. I also own 2 AMD 965 BE systems OC to 3700. My point is that unless AMD really kicks it up a notch they are going to have big trouble with their 8 core

Guskline. That has no bearing on any of this. Again, comparing rated TDP doesn't tell you much of anything as far as performance goes, especially when they don't even measure them in the same manner.

Also, why did you ignore that 4 core Bulldozer has a 95W TDP too? Same number of cores, "same" TDP.

 

[guskline]

Guskline. That has no bearing on any of this. Again, comparing rated TDP doesn't tell you much of anything as far as performance goes, especially when they don't even measure them in the same manner.

Also, why did you ignore that 4 core Bulldozer has a 95W TDP too? Same number of cores, "same" TDP.

Same performance as a 4 core SB?

 

[RussianSensation]

They're not going to ship a slower product.

Prescott was slower than Pentium 4 C at clock speeds up to 3.2ghz. I doubt AMD will flop like that, but it has happened with Intel (also earlier Pentium 4s were slower than Pentium III per clock).

 

[inf64]

Same performance as a 4 core SB?

I expect that 6C 3.6/4.2Ghz Zambezi will be clearly faster than 2500K and positioned against this model(perf. and price wise). I expect that 4C 3.6/3.8Ghz Zambezi will be slower than 2500K by some 15% or so while costing 180$(cheaper than 189$ 1100T while performing practically the same ,on average;much better priced than 2500K which sits at 219$ now).

For 8C I expect that 8C 3.6/4.2Ghz top end model should be around 38% faster than 1100T while costing ~300$ or so.It will be faster than 2600K(7% on average and in line with 980x or 990x) and priced slightly below 2600K,providing better value,as Thuban does now Vs i7 870: 4% slower on average while costing 42% less : 189$ vs 270+$.

 

[guskline]

I expect that 6C 3.6/4.2Ghz Zambezi will be clearly faster than 2500K and positioned against this model(perf. and price wise). I expect that 4C 3.6/3.8Ghz Zambezi will be slower than 2500K by some 15% or so while costing 180$(cheaper than 189$ 1100T while performing practically the same ,on average;much better priced than 2500K which sits at 219$ now).

For 8C I expect that 8C 3.6/4.2Ghz top end model should be around 38% faster than 1100T while costing ~300$ or so.It will be faster than 2600K(7% on average and in line with 980x or 990x) and priced slightly below 2600K,providing better value,as Thuban does now Vs i7 870: 4% slower on average while costing 42% less : 189$ vs 270+$.

Since none of the bulldozers have been released to test, how do you arrive at these calculations?

 

[eternalone]

So much fake performance figures out there it makes me all confused. One minute its groundbreaking the next minute its a flop.

 

[dma0991]

Since none of the bulldozers have been released to test, how do you arrive at these calculations?

Hopefully an 8 core BD will be at least as good if not better than the Core i7 2600K in terms of overall performance. If they are still very very far behind I just might have to wait for Ivy Bridge.

 

[RussianSensation]

dma0991, based on that chart A8 CPUs compete with i3-21xx series but i3-21xx beats a Phenom II X4 in games and office tasks, nevermind the A8. We can't deduce anything performance wise from that AMD chart (you now how it is, internal slides are generally over-exaggerated).

 

[utahraptor]

I remember when the leaks and rumors indicated this card would blow chunks and the shader count was vastly underestimated. I am hoping AMD can recreate this magic by releasing something that surprises everyone.

I know deep down that Intel has a 2550K/2650K ready to be released at a moments notice... Maybe perhaps even on BD release date?

 

[dma0991]

dma0991, based on that chart A8 CPUs compete with i3-21xx series but i3-21xx beats a Phenom II X4 in games and office tasks, nevermind the A8. We can't deduce anything performance wise from that AMD chart (you now how it is, internal slides are generally over-exaggerated).

I'm just showing the product positioning based on price but it's a shame to see that the A6 and A8 is positioned below the Core i3 2100 and the 4 core BD somewhat overlaps with the Core i3 2100. I do not have the intention of building a Llano based rig to begin with and I know how bad it performs since BD cores only comes with Trinity. Hopefully BD is not a major face palm though.

 

[max347]

dma0991, based on that chart A8 CPUs compete with i3-21xx series but i3-21xx beats a Phenom II X4 in games and office tasks, nevermind the A8. We can't deduce anything performance wise from that AMD chart (you now how it is, internal slides are generally over-exaggerated).

That chart shows system cost, and compares them in price- not cpu vs cpu performance. Sure, the i3 beats the p2 as a cpu in pretty much everything, but the chart is showing, that for $x, you can have an i3 rig with intel graphics, or an A8 rig with on die good graphics. That is where they are competing, bigger whole system picture per dollar, not calculating pi on a bench.

 

[Arachnotronic]

Where in the hell is this CPU? I was really hoping to be using the FX chips in some of the builds I've been doing in my little PC biz. Instead I've been building i5 rigs. Good job, AMD.

 

[Davidh373]

There you go :
http://www.chiphell.com/thread-250461-1-1.html

yikes... pretty sad if this is all they have got...

So lets take a look back...

1st: "People" say BD is the Intel Killer
2nd: "People" say BD will be a little better than SB and possibly compete with SB-E
3rd: "People" say BD will compete with regular SB
4th: BD actually ends up worse than Regular SB by about 15%-25% depending on the benchmark...

Why don't people ever learn... I know that these results aren't official or final, but where all the rumors and scattered underground benchmarks have lead me is BD just barely competes with standard SB. That is my expectations now. I mean what has AMD been doing for the last eight months?

 

[RussianSensation]

That chart shows system cost, and compares them in price- not cpu vs cpu performance. Sure, the i3 beats the p2 as a cpu in pretty much everything, but the chart is showing, that for $x, you can have an i3 rig with intel graphics, or an A8 rig with on die good graphics. That is where they are competing, bigger whole system picture per dollar, not calculating pi on a bench.

Ya, for a budget gaming rig, Llano > HD3000 no question. My point was that you can't simply look at corporate slides and assume that the information in them is reflective of actual performance on average. For all we know AMD could have focused on these 3 things with BD:

1) Multi-threaded performance > single-threaded
2) The most emphasis on power consumption to win back notebook and server markets
3) Emphasis on selling more cores for servers

Look at the Projected revenue streams for the firm:
http://www.trefis.com/company#/AMD?from=search

Notebook - 32%
Server - 22%
Desktop - 31%

Realistically speaking, if AMD's BD is a homerun for Notebook and Server market, then it's still a great design. The desktop is a slow growth market, while notebooks and servers is where the most profit margins and growth is.

So really, AMD might have optimized BD for those markets instead. Obviously I am just guessing, but my point is it's not necessary for BD to better than SB on the desktop. As long as AMD hits 2 out of 3 key markets with BD, it's still a winner!

 

[inf64]

Since none of the bulldozers have been released to test, how do you arrive at these calculations?

I have written a blog about this some time ago. Every single bit of how I've got there is detailed .

yikes... pretty sad if this is all they have got...

So lets take a look back...

1st: "People" say BD is the Intel Killer
2nd: "People" say BD will be a little better than SB and possibly compete with SB-E
3rd: "People" say BD will compete with regular SB
4th: BD actually ends up worse than Regular SB by about 15%-25% depending on the benchmark...

Why don't people ever learn... I know that these results aren't official or final, but where all the rumors and scattered underground benchmarks have lead me is BD just barely competes with standard SB. That is my expectations now. I mean what has AMD been doing for the last eight months?

Have your read a word of what I wrote in my post.I practically provided proof that all those chiphell numbers are bogus. It can't cost 300$ and perform on par per clock than age old X4 at the same clock.Forget about Thuban X6,this 8C BD ES is 30% slower according to these benchmarks and will thus cost you 58% more for 30% lower performance.How in the world do you expect this to happen? It won't.

I remember when the leaks and rumors indicated this card would blow chunks and the shader count was vastly underestimated. I am hoping AMD can recreate this magic by releasing something that surprises everyone.

I know deep down that Intel has a 2550K/2650K ready to be released at a moments notice... Maybe perhaps even on BD release date?

Take a look at this desktop performance summary chart. In this chart,on average, 310$ 2600K(3.4/3.8Ghz) is just 7% slower than top of the line 990x with 6 cores and 3.46Ghz/3.7Ghz that costs 900+$. This is solely due to the fact that desktop workload mix is more serial than parallel: many desktop applications don't scale well or scale very poorly past 4 cores/threads. This in turn makes 3.4Ghz SB quad core just barely(7%) slower than 990x.IF you would run a heavily multithreaded workload on both,990x would trounce 2600K by ~40% margin,but the nuber of such applications in desktop world is VERY limited.

So if intel was to release 3.5Ghz+ base clock SB quad core model what do you think this would do to 980x/990x price/performance ? Even with 2600K those two Westmere 6C products make zero sense to even enthusiast OCers,let alone average or above average desktop users. That's right,a hypothetical 2650/2700K SB would completely invalidate the whole high end 6C i7 line and probably even most of the SB-E 6C line since it would be 10-15% or so slower than SB-E 6C @ 3.3Ghz ,on average,while costing 3x less.

This is why 8C Zambezi may end up a hair faster than 2600K on desktop.And if it does it means its single thread performance is quite strong and its multithread performance Vs 2600K would basically be repeat of 990x/980x Vs 2600K.

 

[StinkyPinky]

43 pages of guessing.

 

[Arg Clin]

I have written a blog about this some time ago. Every single bit of how I've got there is detailed .
Have your read a word of what I wrote in my post.I practically provided proof that all those chiphell numbers are bogus. It can't cost 300$ and perform on par per clock than age old X4 at the same clock.Forget about Thuban X6,this 8C BD ES is 30% slower according to these benchmarks and will thus cost you 58% more for 30% lower performance.How in the world do you expect this to happen? It won't.

Very good point! There is an absolute lack of good old common sense to the FUD speculations that BD will not perform competitatively to similar priced intel parts.

If these benches were true, AMD would not bother bringing it to the market.

With that said, I don't know if they are bogus or crippled chips - either way, to present them as any kind of indication of the performance we as consumers can expect from BD is IMO incorrect.