AMD CPU speculation... and expert conjecture

[juanrga]

I hope you have a recipe for crow, cause you're about to be eating it:

http://openbenchmarking.org/prospect/1305170-UT-LLVMCLANG75/fd501a41a2adcc643acc832de94444f9fd7d9678

EDIT: It isn't by a little bit either...the 8350 is like 700% faster...

lol, you gave that benchmark again. I here you say you are a game developer or something like that but give this nonsense out many times really believing it is 7x faster. You gave that benchmark again, the fx8350 has ssd, the i3 3217u which beats the i7 3960x has a hdd with msata ssd I believe and the i7 3960x is tested with a normal hdd. So put in a very fast ssd, a top of the line one and you could destroy the ssd used with the the fx8350.

LOL and more nonsense. Both the FX-8350 and the i7-3960x were tested with an OCZ SSD.

 

[simon12]

Please can we have a new sticky for arguing about power consumption and misinformation so I can find the good and interesting posts on here, that are few and far between.

 

[juanrga]

Hahaha, wow 8350, it seems you need to go back to school:

1. No specific memory model was mentioned, along with no specifications, like speed and per stick size.

2. Different GPUs were used, you may not think, but they can make a difference.

3. Disks are different models completely and different sizes.

4. COMEPLTELY different motherboard brands, at least keep them the same brand.

5. ONLY used the Radeon display driver, why not try an Nvidia? or maybe something else that you can find?

6. different OpenGL Mesas were used.

Me and you know that benchmarks are science experiments, and as I and the United States Education System says, ALL variables in an experiment MUST be the same EXCEPT for ONLY the TESTING MATERIAL (CPU).

And I like how you ONLY put one test of the other 20 tests up. about 70% of the test was won or dominated by Intel, and the 3970X won 40% of that. You forgot really about GraphicsMagick as well.

And you're telling me that I'M giving invalid information? Please, try again with the correct response.

Memory was the same for both and by standard methodology the speed used is stock.

I don't know if this is explained in your United States Education System, but in any experiment there is both relevant and irrelevant variables, experts know which is each one. I can assure you that changing the latencies of the memory modules or the brand of the mobo will not change the score by a 700%.

Also you seem to miss that 8350rocks was mentioning a single benchmark because an ill-informed poster pretended that no benchmark as that existed.

Finally, you idea (5) to test a Nvidia driver in a Radeon card for a CPU test was funny. I will comment it to Michael and maybe in the next benchmark round he will try your scientific suggestion.

 

[cowboy44mag]

It just amazes me how far ahead Intel fans think i7 Haswell is. I have seen many posts on many threads, including a few on this one, where Intel fans claim Steamroller would have to be an 80 - 100% improvement over Piledriver to reach Sandy Bridge levels of preformance. AMD won't catch up to Haswell until 2018, meaning that AMD is 5 years tech wise behind Intel. It is sad that some people actually believe that crap.

If Steamroller was an 80-100% improvement over Piledriver it wouldn't be at Sandy Bridge levels it would dominate Haswell. There is no way that AMD is 5 years behind Intel tech wise, that is just one of the most stupid boasts I've seen. Truth of the matter is Piledriver is knocking on Intel's door right now. It is a fair estimation to put FX 8350 between i5 and i7 levels of performance and in some mulit-threaded applications will actually beat i7. Steamroller is making vast improvements to single core execution, and will improve upon Piledrivers multi-core execution. I'm not saying that Steamroller will dominate i7 Haswell, but Steamroller should put AMD back "on the map". When you look at improvements per generation Sandy Bridge- Ivy Bridge - Haswell isn't anything spectacular, AMD is making much larger improvements per generation Bulldozer - Piledriver - Steamroller. Its not going to take AMD as long to "catch up" as Intel fans think it will. AMD is making big strides forward and Intel is coming across as downright lazy.

 

[cowboy44mag]

Lets get down to the "nuts and bolts" of Intel's Haswell. Some posters on this forum have called Haswell Intel's Bulldozer. That is far too generous. Intel is the industry leader, no if ands or buts. Intel has a HUGE R&D budget. Haswell is FOURTH generation arch. For fourth generation arch every one of the developers working on Haswell should have been fired.

"Performance[edit]
Compared to Ivy Bridge:
Approximately 8% better vector processing performance.[8]
4%-6% IPC (instructions per clock) increase.
3% faster single-threaded performance.
5% faster multi-threaded performance.
Haswell is around 10°C-15°C hotter than Ivy Bridge when overclocked and does not seem to break the 4.4-4.5 GHz mark nearly as easily.
Haswell draws around 30 W more under load than Ivy Bridge.[9][10]
A 4% increase in sequential CPU performance (eight execution ports per core versus six).[11][12]
Up to 25% performance increase over the integrated HD4000 GPU (Haswell HD4600 vs Ivy Bridge's built-in Intel HD4000).[13][14]"

That excerpt was from Wikipedia, even they know what a joke Haswell is compared to Ivy Bridge. So please can all the Intel fanboyism love of Haswell stop!! It isn't a great step forward, and make no mistake with its lackluster performance (especially for a "tock") Intel has left the door wide open for Steamroller.

 

[Cazalan]

Across an entire benchmark suite you might have a point, but this test is pure integer performance...which is done exclusively by the CPU.

Postgre SQL is a open source Database application. How would that correlate to pure integer performance?
http://en.wikipedia.org/wiki/PostgreSQL

Anyhow the reason for the disparity is this benchmark uses FMA which Sandy Bridge does not support. http://openbenchmarking.org/test/pts/pgbench

When you run it on a Haswell CPU the numbers are much different.
http://openbenchmarking.org/result/1306202-SO-UBUNTUHAS15

4727: Haswell Core i7 4770K
2438: AMD FX8350

So that would make Haswell 4 core a whopping 13x faster than Sandy 6 core.

PS: Do some research before posting obscure benchmarks. They might come back to bite you.

 

[The Q6660 Inside]

What opteron beats the i7 3960x
http://www.cpubenchmark.net/high_end_cpus.html
AMD Opteron 6234 12 core cpu loses to laptop cpus, so comparing opteron to the i7 3960x is an insult to intel cpu. I think amd fanboys are so deluded they actually think there cpus are close.

Across an entire benchmark suite you might have a point, but this test is pure integer performance...which is done exclusively by the CPU.

Postgre SQL is a open source Database application. How would that correlate to pure integer performance?
http://en.wikipedia.org/wiki/PostgreSQL

Anyhow the reason for the disparity is this benchmark uses FMA which Sandy Bridge does not support. http://openbenchmarking.org/test/pts/pgbench

When you run it on a Haswell CPU the numbers are much different.
http://openbenchmarking.org/result/1306202-SO-UBUNTUHAS15

4727: Haswell Core i7 4770K
2438: AMD FX8350

So that would make Haswell 4 core a whopping 13x faster than Sandy 6 core.

PS: Do some research before posting obscure benchmarks. They might come back to bite you.

My brother is right on this, that one you linked has a 240gb ocz vertex 3. These benchmarks got virtually nothing to do with the cpu, just ssd or hdd.

Now, shall we get back to AMD Steamroller speculation?

 

[jed]

Across an entire benchmark suite you might have a point, but this test is pure integer performance...which is done exclusively by the CPU.

Postgre SQL is a open source Database application. How would that correlate to pure integer performance?
http://en.wikipedia.org/wiki/PostgreSQL

Anyhow the reason for the disparity is this benchmark uses FMA which Sandy Bridge does not support. http://openbenchmarking.org/test/pts/pgbench

When you run it on a Haswell CPU the numbers are much different.
http://openbenchmarking.org/result/1306202-SO-UBUNTUHAS15

4727: Haswell Core i7 4770K
2438: AMD FX8350

So that would make Haswell 4 core a whopping 13x faster than Sandy 6 core.

PS: Do some research before posting obscure benchmarks. They might come back to bite you.

+1

 

[Cazalan]

Yeah the 8 core haswell 130-140w tdp basically doubling the performance on sime applications like cinebench will smoke any steamroller cpu that will come out. Intel cpus are 2x better performance per watt, quite a big lead I say.
http://www.hardware.fr/medias/photos_news/00/39/IMG0039475.png

The 8 core Haswell will be like $1000 too. Entirely irrelevant to 99.999% of the market.

About as relevant as those people that OC FX-8350 to 8Ghz on LN.

 

[rmpumper]

I hope you have a recipe for crow, cause you're about to be eating it:

http://openbenchmarking.org/prospect/1305170-UT-LLVMCLANG75/fd501a41a2adcc643acc832de94444f9fd7d9678

EDIT: It isn't by a little bit either...the 8350 is like 700% faster...

Good for you. Now tell us how does that 100% irrelevant benchmark (of which 99/100 PC enthusiasts never even heard of) makes the 8350 a better CPU in real life and in real software that people actually use? It doesn't.

 

[rmpumper]

8350- I wouldn't even waste any more effort with this guy. It is clear that he is a total Intel fanboy, and like all Intel fanboys can't believe that an FX 8350 can beat an i7 in ANY benchmark. Intel fanboys are why I would really love to Steamroller come out on totally equal ground vs i7 Haswell in performance and benchmarks. Intel trys to control the information to make their product look better. Not unlike a dictatorship, if the only news you get is gov't approved and they tell you what the gov't wants you to believe you are going to believe it. Intel uses biased benchmarks and review sites to try to control the market. Crysis 3 benchmarks kind of blow the doors off that though as the FX 8350 isn't that far behind the i7 (and in some Crysis 3 benchmarks I've seen beats the i7), and outright beats the i5. Real world benchmarks, like the Crysis 3 benchmarks are what people should be looking out, not synthetic benchmarks designed to give an edge to Intel.

That's coming from a conspiracy theorist AMD fanboy grasping on a single irrelevant benchmark showing that AMD is faster than Intel. Dude, seriously?

 

[griptwister]

Again, this isn't an Intel thread. Go somewhere else and believe what you want. This is a AMD Thread, Not Intel Fanboys rage on AMD enthusiasts. The only ones I see here that are fanboys are the ones that won't shut up about Intel on a AMD thread. Seriously, how many of us on this Thread comment about AMD on the Broadwell speculation and conjuncture?

I found this too.

http://www.eteknix.com/amd-fx-9590-benchmarks-revealed/

 

[jdwii]

Yeah the 8 core haswell 130-140w tdp basically doubling the performance on sime applications like cinebench will smoke any steamroller cpu that will come out. Intel cpus are 2x better performance per watt, quite a big lead I say.
http://www.hardware.fr/medias/photos_news/00/39/IMG0039475.png

The 8 core Haswell will be like $1000 too. Entirely irrelevant to 99.999% of the market.

About as relevant as those people that OC FX-8350 to 8Ghz on LN.

Yes but in 2047 he will be able to say he got the better deal.

 

[noob2222]

Across an entire benchmark suite you might have a point, but this test is pure integer performance...which is done exclusively by the CPU.

Postgre SQL is a open source Database application. How would that correlate to pure integer performance?
http://en.wikipedia.org/wiki/PostgreSQL

Anyhow the reason for the disparity is this benchmark uses FMA which Sandy Bridge does not support. http://openbenchmarking.org/test/pts/pgbench

When you run it on a Haswell CPU the numbers are much different.
http://openbenchmarking.org/result/1306202-SO-UBUNTUHAS15

4727: Haswell Core i7 4770K
2438: AMD FX8350

So that would make Haswell 4 core a whopping 13x faster than Sandy 6 core.

PS: Do some research before posting obscure benchmarks. They might come back to bite you.

interesting that FMA speeds up that fast...
so basically this would be like running with certain features disabled on a cpu ... sound familiar? its called cinebench ... ICC anyone? ya, we already been there except the shoe is on the other foot.

 

[juanrga]

It just amazes me how far ahead Intel fans think i7 Haswell is. I have seen many posts on many threads, including a few on this one, where Intel fans claim Steamroller would have to be an 80 - 100% improvement over Piledriver to reach Sandy Bridge levels of preformance. AMD won't catch up to Haswell until 2018, meaning that AMD is 5 years tech wise behind Intel. It is sad that some people actually believe that crap.

If Steamroller was an 80-100% improvement over Piledriver it wouldn't be at Sandy Bridge levels it would dominate Haswell. There is no way that AMD is 5 years behind Intel tech wise, that is just one of the most stupid boasts I've seen. Truth of the matter is Piledriver is knocking on Intel's door right now. It is a fair estimation to put FX 8350 between i5 and i7 levels of performance and in some mulit-threaded applications will actually beat i7. Steamroller is making vast improvements to single core execution, and will improve upon Piledrivers multi-core execution. I'm not saying that Steamroller will dominate i7 Haswell, but Steamroller should put AMD back "on the map". When you look at improvements per generation Sandy Bridge- Ivy Bridge - Haswell isn't anything spectacular, AMD is making much larger improvements per generation Bulldozer - Piledriver - Steamroller. Its not going to take AMD as long to "catch up" as Intel fans think it will. AMD is making big strides forward and Intel is coming across as downright lazy.

I am not sure if them believe that crap or are paid for spreading it in forums and blogs.

 

[juanrga]

Across an entire benchmark suite you might have a point, but this test is pure integer performance...which is done exclusively by the CPU.

Postgre SQL is a open source Database application. How would that correlate to pure integer performance?
http://en.wikipedia.org/wiki/PostgreSQL

Anyhow the reason for the disparity is this benchmark uses FMA which Sandy Bridge does not support. http://openbenchmarking.org/test/pts/pgbench

When you run it on a Haswell CPU the numbers are much different.
http://openbenchmarking.org/result/1306202-SO-UBUNTUHAS15

4727: Haswell Core i7 4770K
2438: AMD FX8350

So that would make Haswell 4 core a whopping 13x faster than Sandy 6 core.

PS: Do some research before posting obscure benchmarks. They might come back to bite you.

interesting that FMA speeds up that fast...
so basically this would be like running with certain features disabled on a cpu ... sound familiar? its called cinebench ... ICC anyone? ya, we already been there except the shoe is on the other foot.

Apples and oranges. The Haswell score shows that the benchmark is not disabling anything, merely that the sandy bridge chip lacks FMA support. Running a test that uses all the possibilities of the chips is fair. If a given chip lacks some instruction set or technology then will run slower. That is the goal of a benchmark.

To be comparable to ICC-sysmark-cinebench the test would use CPUID to detect the Haswell chip and then force it to not run the FMA way, even if the chip is compatible with it.

As said apples and oranges. Or fair vs unfair.

 

[gamerk316]

Hahaha, wow 8350, it seems you need to go back to school:

1. No specific memory model was mentioned, along with no specifications, like speed and per stick size.

2. Different GPUs were used, you may not think, but they can make a difference.

3. Disks are different models completely and different sizes.

4. COMEPLTELY different motherboard brands, at least keep them the same brand.

5. ONLY used the Radeon display driver, why not try an Nvidia? or maybe something else that you can find?

6. different OpenGL Mesas were used.

Me and you know that benchmarks are science experiments, and as I and the United States Education System says, ALL variables in an experiment MUST be the same EXCEPT for ONLY the TESTING MATERIAL (CPU).

And I like how you ONLY put one test of the other 20 tests up. about 70% of the test was won or dominated by Intel, and the 3970X won 40% of that. You forgot really about GraphicsMagick as well.

And you're telling me that I'M giving invalid information? Please, try again with the correct response.

Memory was the same for both and by standard methodology the speed used is stock.

I don't know if this is explained in your United States Education System, but in any experiment there is both relevant and irrelevant variables, experts know which is each one. I can assure you that changing the latencies of the memory modules or the brand of the mobo will not change the score by a 700%.

Also you seem to miss that 8350rocks was mentioning a single benchmark because an ill-informed poster pretended that no benchmark as that existed.

Finally, you idea (5) to test a Nvidia driver in a Radeon card for a CPU test was funny. I will comment it to Michael and maybe in the next benchmark round he will try your scientific suggestion.

Across an entire benchmark suite you might have a point, but this test is pure integer performance...which is done exclusively by the CPU.

Postgre SQL is a open source Database application. How would that correlate to pure integer performance?
http://en.wikipedia.org/wiki/PostgreSQL

Anyhow the reason for the disparity is this benchmark uses FMA which Sandy Bridge does not support. http://openbenchmarking.org/test/pts/pgbench

When you run it on a Haswell CPU the numbers are much different.
http://openbenchmarking.org/result/1306202-SO-UBUNTUHAS15

4727: Haswell Core i7 4770K
2438: AMD FX8350

So that would make Haswell 4 core a whopping 13x faster than Sandy 6 core.

PS: Do some research before posting obscure benchmarks. They might come back to bite you.

Across an entire benchmark suite you might have a point, but this test is pure integer performance...which is done exclusively by the CPU.

Postgre SQL is a open source Database application. How would that correlate to pure integer performance?
http://en.wikipedia.org/wiki/PostgreSQL

Anyhow the reason for the disparity is this benchmark uses FMA which Sandy Bridge does not support. http://openbenchmarking.org/test/pts/pgbench

When you run it on a Haswell CPU the numbers are much different.
http://openbenchmarking.org/result/1306202-SO-UBUNTUHAS15

4727: Haswell Core i7 4770K
2438: AMD FX8350

So that would make Haswell 4 core a whopping 13x faster than Sandy 6 core.

PS: Do some research before posting obscure benchmarks. They might come back to bite you.

interesting that FMA speeds up that fast...
so basically this would be like running with certain features disabled on a cpu ... sound familiar? its called cinebench ... ICC anyone? ya, we already been there except the shoe is on the other foot.

Apples and oranges. The Haswell score shows that the benchmark is not disabling anything, merely that the sandy bridge chip lacks FMA support. Running a test that uses all the possibilities of the chips is fair. If a given chip lacks some instruction set or technology then will run slower. That is the goal of a benchmark.

To be comparable to ICC-sysmark-cinebench the test would use CPUID to detect the Haswell chip and then force it to not run the FMA way, even if the chip is compatible with it.

As said apples and oranges. Or fair vs unfair.

So in conclusion, based on 8350's argument, Haswell crushes PD.

So yeah, understanding numbers is important. Its not good enough to say "8350 is 700x faster in a benchmark!", you also have to answer "why is the 8350 700x faster in a benchmark". In this case: FMA support, which I think was added in IB.

As a general rule, since SSE3 is still the lowest supported common set of CPU instructions, if making a benchmark software for the purposes of performance comparisons, benchmarks should probably be compiled against SSE3. The effects of various CPU optimizations should be looked at separately.

 

[8350rocks]

Across an entire benchmark suite you might have a point, but this test is pure integer performance...which is done exclusively by the CPU.

Postgre SQL is a open source Database application. How would that correlate to pure integer performance?
http://en.wikipedia.org/wiki/PostgreSQL

Anyhow the reason for the disparity is this benchmark uses FMA which Sandy Bridge does not support. http://openbenchmarking.org/test/pts/pgbench

When you run it on a Haswell CPU the numbers are much different.
http://openbenchmarking.org/result/1306202-SO-UBUNTUHAS15

4727: Haswell Core i7 4770K
2438: AMD FX8350

So that would make Haswell 4 core a whopping 13x faster than Sandy 6 core.

PS: Do some research before posting obscure benchmarks. They might come back to bite you.

Well, I am glad you asked...

Database functions and indexing require purely integer specific functions. As you process a database you are constantly using Add/Multiply functions to crunch the database and convert the information or adjust it, etc. The functions that Postgre SQL uses require no floating point calculations...

Anytime you are using a database you are using pure integer performance...floating point never comes into the equation. So, in essence, as I said...it is only a measure of pure integer performance. I put it into extremely simple terms that he would understand, however, I was not wrong...and hence, it has not come back to "bite me". I deal with databases pretty regularly...when you design a game, and skills/talents/powers/abilities are made, then stored in a database, and when that "power" is accessed, a pointer to the database library is pulled up telling the system which part of the database to execute in order to operate that power.

Additionally...the fact that AMD includes more instructions on their CPUs than Intel does has no bearing on the CPUs capability. It just means Intel didn't build that functionality into their CPU...if the 8350 didn't have AVX code support, and was performing poorly in an AVX benchmark...Intel fanboys would crow from the rooftops about how terrible it was...even though the code was simply unsupported.

So, next time, remember who you're talking to.



EDIT: Postgre SQL data types/functions (per wiki):

Data types[edit]

A wide variety of native data types are supported, including:
Boolean
Arbitrary precision numerics
Character (text, varchar, char)
Binary
Date/time (timestamp/time with/without timezone, date, interval)
Money
Enum
Bit strings
Text search type
Composite
Arrays (variable length and can be of any data type, including text and composite types) up to 1 GB in total storage size.
Geometric primitives
IPv4 and IPv6 addresses
CIDR blocks and MAC addresses
XML supporting XPath queries
UUID
JSON (versions 9.2 and up)

In addition, users can create their own data types which can usually be made fully indexable via PostgreSQL's GiST infrastructure. Examples of these include the geographic information system (GIS) data types from the PostGIS project for PostgreSQL.

There is also a data type called a "domain", which is the same as any other data type but with optional constraints defined by the creator of that domain. This means any data entered into a column using the domain will have to conform to whichever constraints were defined as part of the domain.

Range Types[edit]

Starting with PostgreSQL 9.2, a data type that represents a range of data can be used which are called range types. These can be discrete ranges (e.g. all integer values 1 to 10) or continuous ranges (e.g. any point in time between 10:00am and 11:00am). The built-in range types available include ranges of integers, big integers, decimal numbers, time stamps (with and without time zone) and dates.

Custom range types can be created to make new types of ranges available, such as IP address ranges using the inet type as a base, or float ranges using the float data type as a base. Range types support inclusive and exclusive range boundaries using the [] and () characters respectively. (e.g. '[4,9)' represents all integers starting from and including 4 up to but not including 9.) Range types are also compatible with existing operators used to check for overlap, containment, right of etc.

 

[cowboy44mag]

So basically this AMD thread has now officially become a Haswell is the greatest thing every invented by mankind ever, ever, ever, ever, ever. We should throw ourselves at Intel's feet and worship their glorious processor as god. Is that about right?

I mean lets just ignore the fact that almost anywhere you look all you find is reports that Haswell isn't as good as Ivy Bridge, but lets ignore all those facts because Intel has become its very own religion and can do no wrong.

Can there be one thread about AMD on the internet that isn't hijacked by a bunch of over-privileged guys that think because they had much larger budgets to work with they are soooo much better and smarter than the rest of us? I mean we got guys on this thread bragging about how a $1000 Intel processor will crush anything AMD can bring, really ~$1000 is usually the entire budget that most people building a computer system have. But here come the over-privileged elitist bragging about how many i7 computers they currently have and chomping at the bit to get ahold of an 8 core Haswell. Yea, we get it you can afford the Lamborghini and the rest of us are stuck with economy cars. Kudos to you for having so much money that you have no idea what to do with it next. Simply buying the most expensive items you can find doesn't make you more intelligent, making the value oriented brand (price/performance) run nearly as good as the overpriced brand is much more impressive. Example I needed something a little bigger than an F-350 for my business, but really didn't want to pay the bloated price of an F-550. Added some leaf springs (front and back), an airbag lift kit, some engine tweaks and alterations... my F-350 has nearly all the same stats as an F-550 for a fraction the cost (got to admit a lot of work though), on top of that I'm able to get it through inspection every year and register it every year at the cheaper F-350 rates. However Intel fans are sooo impressed with the brandname they wouldn't look any further than it says F-350 so it must be inferior to the F-550, just like they have all passed judgement on Steamroller as being vastly inferior to Haswell because herp a derp it don't say Intel Inside anywhere on it. AMD's current Piledriver FX-8350 can get some impressive results but you have to put a little effort into tweaking and overclocking it to get it there, and Steamroller will bring AMD even further.

/end rant

For the rest of us, who live in the real world and are looking at $200-300 processors and top end motherboards $200 is there actually any Steamroller news? Its getting tiring coming here to get some tid-bit of what Steamroller will offer only to read page after page of Haswell news from Intel fanboys on an AMD thread.

 

[8350rocks]

If you can wade through the BS...I posted some great news yesterday:

HD 9970 due in Oct 2013 to coincide with BF4 launch and a new gaming bundle

FX 9590 benchmarks have supposedly been spotted...

Kaveri low power Engineering Samples were spotted in a data base...the results are very early...so there is no way to extrapolate anything useful (to us) at the moment.

The first FM2+ MB from Asus has already been spotted...meaning Kaveri is coming soon.

 

[cowboy44mag]

If you can wade through the BS...I posted some great news yesterday:

HD 9970 due in Oct 2013 to coincide with BF4 launch and a new gaming bundle

FX 9590 benchmarks have supposedly been spotted...

Kaveri low power Engineering Samples were spotted in a data base...the results are very early...so there is no way to extrapolate anything useful (to us) at the moment.

The first FM2+ MB from Asus has already been spotted...meaning Kaveri is coming soon.

Thanks 8350 I'll have to go back, and as you say wade through the BS to find the actual AMD news. Looking forward to reviewing the FX-9590 benchmarks. Its getting harder and harder waiting for Steamroller FX I doubt the wife would let me shell out the money for an FX 9590 though- have to answer to the boss you know

 

[Cazalan]

Anytime you are using a database you are using pure integer performance...floating point never comes into the equation. So, in essence, as I said...it is only a measure of pure integer performance. I put it into extremely simple terms that he would understand, however, I was not wrong...and hence, it has not come back to "bite me". I deal with databases pretty regularly...when you design a game, and skills/talents/powers/abilities are made, then stored in a database, and when that "power" is accessed, a pointer to the database library is pulled up telling the system which part of the database to execute in order to operate that power.

After looking at more results the FMA dependency looks premature. There's a Kabini (w/FMA) scoring 118 pts and a Trinity APU (w/FMA) scoring 264 pts, and a MacBook Pro (Nehalem i5-520m, no FMA) with 1880 pts. Thats pointing to the I/O subsystem as the bottleneck. SSD drives having much higher scores.

This test would have to be run from a RamDisk to focus purely on CPU performance.

 

[8350rocks]

Anytime you are using a database you are using pure integer performance...floating point never comes into the equation. So, in essence, as I said...it is only a measure of pure integer performance. I put it into extremely simple terms that he would understand, however, I was not wrong...and hence, it has not come back to "bite me". I deal with databases pretty regularly...when you design a game, and skills/talents/powers/abilities are made, then stored in a database, and when that "power" is accessed, a pointer to the database library is pulled up telling the system which part of the database to execute in order to operate that power.

After looking at more results the FMA dependency looks premature. There's a Kabini (w/FMA) scoring 118 pts and a Trinity APU (w/FMA) scoring 264 pts, and a MacBook Pro (Nehalem i5-520m, no FMA) with 1880 pts. Thats pointing to the I/O subsystem as the bottleneck. SSD drives having much higher scores.

This test would have to be run from a RamDisk to focus purely on CPU performance.

The curious thing there is that both systems tested used 60 GB OCZ SSDs...

 

[The Q6660 Inside]

Yeah the 8 core haswell 130-140w tdp basically doubling the performance on sime applications like cinebench will smoke any steamroller cpu that will come out. Intel cpus are 2x better performance per watt, quite a big lead I say.
http://www.hardware.fr/medias/photos_news/00/39/IMG0039475.png

The 8 core Haswell will be like $1000 too. Entirely irrelevant to 99.999% of the market.

About as relevant as those people that OC FX-8350 to 8Ghz on LN.

Yes but in 2047 he will be able to say he got the better deal.

Not to mention, 140W TDP=trololollolololololololololol, you would need at least 190W, considering the increases in various parts of the CPU.

 

[The Q6660 Inside]

Anytime you are using a database you are using pure integer performance...floating point never comes into the equation. So, in essence, as I said...it is only a measure of pure integer performance. I put it into extremely simple terms that he would understand, however, I was not wrong...and hence, it has not come back to "bite me". I deal with databases pretty regularly...when you design a game, and skills/talents/powers/abilities are made, then stored in a database, and when that "power" is accessed, a pointer to the database library is pulled up telling the system which part of the database to execute in order to operate that power.

After looking at more results the FMA dependency looks premature. There's a Kabini (w/FMA) scoring 118 pts and a Trinity APU (w/FMA) scoring 264 pts, and a MacBook Pro (Nehalem i5-520m, no FMA) with 1880 pts. Thats pointing to the I/O subsystem as the bottleneck. SSD drives having much higher scores.

This test would have to be run from a RamDisk to focus purely on CPU performance.

The curious thing there is that both systems tested used 60 GB OCZ SSDs...

The AMD had the advantage of a Vertex 2, but that was a somewhat minor increase over the Vertex used in the Intel.