Official Phenom Reviews Thread

[freeloader]

http://www.newegg.com/Product/Produc...82E16819103226

I will buy one when they hit ~220-230.
It's not the best CPU in the world, but I love it.

A B3 revision would be really nice at that price.

[JumpingJack]

Realy ? lets take a look at some multicore affected Benchmarks

Well there is

here
http://img213.imageshack.us/img213/5770/bioshockky2.gif

and here
http://img339.imageshack.us/img339/3...netcaveer2.gif

here again
http://img402.imageshack.us/img402/1788/divxvx0.gif

oh here is another on
http://img212.imageshack.us/img212/5...conceptrj6.png

next one here
http://img442.imageshack.us/img442/1...vchess2ug5.gif

hell there a alot
http://img337.imageshack.us/img337/3...erendersc4.gif

Want more?

Well, for another example, if you took all the multicore optimized benches in the THG Benchmark together , there is a gap of 1 % between GP-9600 and Q6600.

Again i don't consider any benches which takes no benefit of the mutlticore Design.

Yes, Mainconcept as well as Windows Media Encoder showed good results with Phenom, unforunately that was about it... The Lost Planet demo you chose is interesting data, but other sites are showing the same demo (not GPU limited) about even or getting beat clock for clock (the 9900 is about the same a Q6600):
http://www.firingsquad.com/hardware/...view/page5.asp

While you picked thorugh each review and chose select graphs that suit your argument, the overall trend with the entire body of data is not as good as you would make it out to be.... this is called 'selective data analysis', that's ok... frankly, we will soon have more data as the press moves away from the Tahoe conditions to in lab conditions, in which it may actually improve... or it may get worse, we won't know unless the press decides to do follow up reviews.

We certainly get that impression from such places as Firingsquad and other who complained about the short time with systems that they could not build nor had control over.

[duploxxx]

No they run in our corporate production environment. And the dual FSB dont seem to be much of a bottleneck...

You sound desperate these days....

nope not desperate at all, as i said, it is finally a pleasure to work on these Xeons, not like the years before.

buying server should be depending on type of applications, it are fools that control server rooms stating that any system will serve them right as long as it has a name.... but offcourse many have no idea that a cpu/architecture can have a great influence on there sw.

[justapost]

What about sse4a? I found visual studio 2008 will make use of it. This version was released two days ago, according to this news.
I think none of the benchmarks used made use of it.
GCC supports sse4a since version 4.1.2 but i'm not aware of an review covering sse4a benefits.
I guess the impact will not be huge, it are just four new commands but it had not been reviewed.

About pricing

Source

[Shintai]

What about sse4a? I found visual studio 2008 will make use of it. This version was released two days ago, according to this news.
I think none of the benchmarks used made use of it.
GCC supports sse4a since version 4.1.2 but i'm not aware of an review covering sse4a benefits.
I guess the impact will not be huge, it are just four new commands but it had not been reviewed.

About pricing

Source

With all respect and short. SSE4 would yield even more. SSE4A is only 4 instructions. SSE4 is 54.

Also Microsofts compiler is slower than Intels. Both for AMD and Intel CPUs. Even MS themselves uses Intel.

The applications that would benefit from SSE4A would mainly be media stream/conversion for bypassing caches with direct main memory access. Aka encoding.

But SSE4A is more PR than real use. Same reason AMD announced SSE5...

[justapost]

In the article you linked there is no word mentioned about AMD or SSE4A or K10 or anything related to CPUs!

Here you go.

AMD Highlights Optimized Integration between Quad-Core AMD Opteron™ Processors and Upcoming Microsoft Products

— At Microsoft TechEd Developers conference, AMD demonstrates commitment to software developers through Quad-Core AMD Opteron processors and Visual Studio 2008 integration—

BARCELONA, Spain -- November 5, 2007 --AMD (NYSE: AMD) today announced it is demonstrating platforms based on Quad-Core AMD Opteron™ processors, including features designed to help Microsoft Windows application developers optimize their software applications through tools in upcoming software products from Microsoft Corp. The technology demonstrations at the conference in Barcelona, Spain show new integration with Microsoft Visual Studio 2008 and Quad-Core AMD Opteron processors.

Ongoing collaboration throughout product development has ensured that Visual Studio 2008 tools will produce optimized code when compiling for Quad-Core AMD Opteron processors. This hardware/software collaboration is designed to ensure that software developers will enjoy improved instruction selection, optimized register allocation and enhanced 128-bit floating-point performance. “The Microsoft Visual Studio suite represents key building blocks in the Windows software development tools stack and is widely used by the Windows development community. This is why Visual Studio support for AMD processors is so important to us- it provides a highly productive environment for our customers to connect their hardware and software,” said Earl Stahl, vice president of Software Development at AMD. “The hardware/software collaboration is vital across the entire Microsoft Application Platform, so we are also working to ensure that Microsoft Windows Server 2008 and Microsoft SQL Server 2008 are optimized for Quad-Core AMD Opteron processors.”

Engineers from AMD and Microsoft worked side by side to ensure Visual C++ would have code generation improvements for Quad-Core AMD Opteron processors. Other integration improvements include SSE4a and Advanced Bit Manipulation (ABM) intrinsic functions included in Visual Studio 2008 and optimum code generation for Quad-Core AMD Opteron processors.

“Microsoft and AMD share a common goal of providing software developers with the tools they need to be successful. We’ve worked together to establish the software and hardware link between Visual Studio 2008 and Quad-Core AMD Opteron,” said S. Somasegar, corporate vice president of the Developer Division at Microsoft Corporation. “This hardware/software combination provides tools that mainstream software developers need to do their jobs more efficiently to meet the demands of today’s global computing community.”

Onsite at the TechEd Developers Conference, AMD will host two sessions designed to provide developers with tips to improve x86 and x64 performance with native code when using Visual Studio 2008 as well as best practices to fully utilize multi-core processing in managed code. TechEd attendees can visit AMD booth #A-1. For coding tips, sample code and downloads of software development tools and libraries, please visit http://developer.amd.com

About AMD
Advanced Micro Devices (NYSE: AMD) is a leading global provider of innovative processing solutions in the computing, graphics and consumer electronics markets. AMD is dedicated to driving open innovation, choice and industry growth by delivering superior customer-centric solutions that empower consumers and businesses wwww.amd.comorldwide. For more information, visit www.amd.com.

AMD, the AMD Arrow logo, AMD Opteron and combinations thereof, are trademarks of Advanced Micro Devices, Inc. Other names are for informational purposes only and may be trademarks of their respective owners.

Source

GCC 4.2 is intended as a performance-optimized, optional compiler that, in addition to containing the above features, also includes the below enhancements. Click here to download a copy:

* C++ performance bug fixes (benefits AMD APL libraries)
* Improved performance tuning for Family 10h
* Vectorizer improvements
* Minor bug fixes and typographical errors
* Source code cleanup

* » GCC 4.1.2 Downloads

GCC 4.1.2 is an upgrade to the version of GCC shipped with RHEL5 and SLES10 SP1, and adds the following additional features:

* Support for SSE4a instructions introduced in the code name "Barcelona" family of AMD processors
* Support of Advanced Bit Manipulation instructions

http://gcc.gnu.org/wiki/AMDFAM10

[justapost]

Thanks If you posted this link in your previous post I wasn't going to respond like that. As I said, there is no single word about AMD, CPUS, K10 or SSE4A in the article you linked.

I did a quick search after your respond, had info from blogs and newsgroups.
My post was about sse4a not yet reviewed. It might have a little impact on codecs, but no 40% improvement over all.

Even MS themselves uses Intel.

What compiler does MS use for their operating system? AMD suggest MS compiler.
I wonder what a benefit linux distributions will have on k10 arch from gcc improvements as those also tune for prefetching.

[Shintai]

I did a quick search after your respond, had info from blogs and newsgroups.
My post was about sse4a not yet reviewed. It might have a little impact on codecs, but no 40% improvement over all.

What compiler does MS use for their operating system? AMD suggest MS compiler.
I wonder what a benefit linux distributions will have on k10 arch from gcc improvements as those also tune for prefetching.

As I said..Intel compiler...why? Because it gives the greatest benefit.

Why do MS make a compiler for VS2008? Because of cost...

I guess you are on a dreampath somehow to hope SSE4A "fixes" it. Yet if you compiler for SS4A and SSE4. It would just be even worse. SSE4A is mainly a gimmick to pretend to have SSE4 for joe average.

Also AMD dont suggest MS compiler. They even themselves use Intels compiler for Barcelona spec benches.

So with all respect..leave it as it is. We dont need another "wait and see" fantasy. Simply because you eat marketing raw! Do I even have to remind you it was the SAME marketing department telling you barcelona would be 40-50% faster than Core 2 clock for clock?

[justapost]

As I said..Intel compiler...why? Because it gives the greatest benefit.

You have a source for that?

I guess you are on a dreampath somehow to hope SSE4A "fixes" it. Yet if you compiler for SS4A and SSE4. It would just be even worse. SSE4A is mainly a gimmick to pretend to have SSE4 for joe average.

Also AMD dont suggest MS compiler. They even themselves use Intels compiler for Barcelona spec benches.

So with all respect..leave it as it is. We dont need another "wait and see" fantasy. Simply because you eat marketing raw! Do I even have to remind you it was the SAME marketing department telling you barcelona would be 40-50% faster than Core 2 clock for clock?

No need to wait Windows Server 2008 Release Candidate Trial can be used for a review.

[freeloader]

I'd like to see how the Phenom compares to C2Q in a 64 bit environment at the same clock speeds. We all know that C2 is excellent at 32 bit but loses a bit of steam in 64 bit mode. Anyone know of any posted benches where both processors are strictly operating in a 64 bit environment?

[Shintai]

You have a source for that?

No need to wait Windows Server 2008 Release Candidate Trial can be used for a review.

Why would Microsoft use a compiler thats quite abit slower...to help OSX and *nix? I only have access directly to MS coders for debugging. And they should freaking know...

Ok, I guess we could go on forever and ever with you hoping.

Let me give you a hint...check the OS "speedbonus" from NT4 to Windows 2003. Also yet anotehr bigger hint. Vista and 2008 server uses the SAME CODE BASE.

The OS would benefit about 0, zip, nada. SSE is FP. Hardly something the OS itself uses.

But I guess looking on your post that you want to try and hope on another hype.

Have fun with your marketing...I´m done with argueing with dreamers.

[Donnie27]

That was a personal experience note I made that they clarified after, long before now. I was told on September 8th as I asked then, not that they were told on September 8th. To many others, yes AMD withheld all the facts of their product performance => the exaggerated word for that would be "deceived", but I've already said this more than once. They didn't lie to the person I talked to at that government corp, they had already shown them the system up and running since July for demo's with its expected performance and clarified very clearly the "yet to be approved" ballpark figures for a 2.0GHz Barcelona and a 2.6GHz Barcelona in SPECfp_rate and SPECint_rate (2006). Those figures turned out to be fully true. They also said the major improvement was in SPECfp_rate 2006, so they weren't exactly lying to that corp. If they had, the corp wouldn't have looked back at them twice which includes government IT orders and that would've been a heavy loss for AMD.

Sept 8 has or had NOTHING to do with AMD lying and purposely trying to mislead both customers and the General Public in Jan and Feb 07. Again KTE, I RESPECTFULLY disagree with you here. There wasn't any careful wording and the two quotes clearly show AMD talking about blowing away and etc.... They didn't leave themselves any wiggle room at all. Just as picking a name to mean Phenominal performance LOL! As coldpower27, savantu, Shintai, terrance, and many other pointed out. K8 was already faster at SPECfp-rate 2006 and even that was deception and BS on AMD's part. No, K10 is not 3.6 times faster than K8 running SPECfp-rate=P Still a fat lie there as well.

Corporations don't rely on being spoon-fed their data like other customers, they make their analysis from what they are shown early on before release and leave it up to the seller to prove their product to them after that. If they can't, its their loss. They 'aint weak minded enough to fall for "40% improvement over Clovertown" from a well dressed PR woman until they see it and this goes for market analysts too. When the IT department of the hospital I work at asked AMD about what was the fastest Family K10h they can retail "optimistically" by mid-January in early September, they replied very clearly "2.6GHz B2 step is looking like the best possible around that time period". There's no lies in that either and yet all the reason to do so.

WOW, I work at a Hospital as well LOL! Our IT Dept are known to not Beta Test for AMD or Intel. Barkie will be on the market for about 18 moths before it is even given a test run. Yes, we're large enough that Dell and Hp have been know to send EVAL units.

I don't know what the AMD folks from Dell said. I did hear about what they said about Clovertown. Simply, after testing both, Clovertown is faster=P

There's no misunderstanding, I said the same thing before your reply to me. I don't care what AMD or Intel did more than I do that a rat is dying right now DownTown 5th street alley way because screaming about it long after its happened doesn't do anything to curb it nor change it. Would've made a due difference back then though. We can all read what they claimed and what came to be of it. Was I deceived by the 40% marketing figures or what they applied to? No, neither were many others. Who was deceived? Anyone gullible or not understanding marketing rigmarole and computer architecture enough to fall for it. Marketing equals boast of your best facts and hiding of your worst facts, and we knew the SPECfp_rate is where AMD shines most and will boast most about at any given chance. I've already made it clear that it was physically impossible for them to validly claim and for someone educated in computing architecture to accept 40% improvement overall over Clovertown by their transistor technology. All it takes is for someone to understand microprocessors to be able to guess that at BEST K10h could've been a 10% improvement over Core 2, nothing more, which it wasn't. Were other people deceived? Yes. Do I care? No, but why should I? They should learn micrprocessors better before following some salesman's claims blindly, and that goes in anything. It wasn't me or any close one to me who made any loss, nor anyone relying on me.

AND FWIW, you will see a similar trend with the new architecture too. AMD with IBM are working on low-k dielectric. Low-k allows very good power savings, more than high-k but no where near as good clock frequencies as high-k dielectric like Intel chose. At higher clocks the gate leakage starts increasing rapidly. It's one of its major weaknesses.

It is disappointing that AMD has nothing to directly compete clock-for-clock with Penryn though, just like C2D vs X2 basically.

With all that said, AMD still CLEARLY lied and that's all I said. I didn't say I hate AMD and if Phenom did Kick ass, I'd be on it like a Hungry Fat guy on a ham. Phonem is weak, I'm not caring anything about a bright side. I've aways swung my purchases as AMD and Intel traded having the best. Still have two AMD and Two Intel based rigs right now.

Even at 3GHz the Phenom would still be slower than Intel's equivalent. I believe AMD knew this when the first working model was up and running.
BTW, Happy Thanksgiving!

[Face]

Few more:

http://pc.watch.impress.co.jp/docs/2.../tawada118.htm (Japanese, but with english charts)

http://www.nowpug.com/bbs/view.php?id=pug_review&no=75 (Ditto)

http://www.custompc.co.uk/news/60168...rks/page2.html

Edit: btw, here's the first retail Phenom 9500 on XS.

[KTE]

Sept 8 has or had NOTHING to do with AMD lying and purposely trying to mislead both customers and the General Public in Jan and Feb 07. Again KTE, I RESPECTFULLY disagree with you here. There wasn't any careful wording and the two quotes clearly show AMD talking about blowing away and etc....

Actually, you agree with me rather than disagreeing bro. Whatever you said about AMD not making it clear and trying to deceive people into thinking 40% performance over Intel is what they did, and I agreed right from the word go. All I said is, I wasn't deceived neither did I believe in that because it was quite obvious a market tactic to buy time, attention, popularity and drop Intel sales for me.

WOW, I work at a Hospital as well LOL! Our IT Dept are known to not Beta Test for AMD or Intel. Barkie will be on the market for about 18 moths before it is even given a test run. Yes, we're large enough that Dell and Hp have been know to send EVAL units.

I'm only a trainee yet -> part-time. But the department is optical surgery and since its government owned and funded they tend to get contacted by company reps often. I saw this with AMD/Intel myself near to Barcelona launch although they don't contact me simply because I don't have the necessary position.

Still have two AMD and Two Intel based rigs right now.

LOL! That's exactly like me here, how strange Donnie.

Even at 3GHz the Phenom would still be slower than Intel's equivalent. I believe AMD knew this when the first working model was up and running.

Believe me, they would definitely have known by mid-January 07, because their marketing has to be based off the directives received from initial lab testing.

BTW, Happy Thanksgiving!

You too mate.

Edit: btw, here's the first retail Phenom 9500 on XS.

Kazgirl? 'Aint she from AMD forums? I read she was after a Phenom a bit ago there, yeah, I'm sure it was her. On XS?

Wow.

[Donnie27]

Few more:

http://pc.watch.impress.co.jp/docs/2.../tawada118.htm (Japanese, but with english charts)

http://www.nowpug.com/bbs/view.php?id=pug_review&no=75 (Ditto)

http://www.custompc.co.uk/news/60168...rks/page2.html

Edit: btw, here's the first retail Phenom 9500 on XS.

Ouch!

[justapost]

Why would Microsoft use a compiler thats quite abit slower...to help OSX and *nix? I only have access directly to MS coders for debugging. And they should freaking know...

It sounds reasonable, but I could not find any source neighter on intel nor on the microsoft site.
Linux distributions however use gcc.

But I guess looking on your post that you want to try and hope on another hype.

Have fun with your marketing...I´m done with argueing with dreamers.

Well Penry reviews showed the benefit from sse4, i want to see a review about sse4a. If there is no benefit, bad for AMD.

[Sagart]

Why is the 9700 and 9900 so much better when the frequences are just a bit higher? (It seems more effective)

[Shintai]

Why is the 9700 and 9900 so much better when the frequences are just a bit higher? (It seems more effective)

More marketing I guess. Same company that posted they would be 40-50% faster than Core 2...

Also "projected" can cover alot.

The 9500 vs 9600 is a..."miscalculation"...aswell. Its 2.2Ghz vs 2.3Ghz. They say 7% more performance. Yet even with 100% scaling. The core difference is only 4.5%.

[JumpingJack]

Well Penry reviews showed the benefit from sse4, i want to see a review about sse4a. If there is no benefit, bad for AMD.

SSE4a consists of 4 instructions, none of the 4 are related to Intel's SSE4 instruction set. I have not seen any data showing any information on SSE4a compiled software vs non-SSE4a compiled software.... but a resonable assumption is that it does not provide a huge improvement otherwise AMD would have advertised it as a significant selling point.

http://developer.amd.com/assets/Deve...in_Boggs-2.pdf
Page 17

[nemrod]

The 9500 vs 9600 is a..."miscalculation"...aswell. Its 2.2Ghz vs 2.3Ghz. They say 7% more performance. Yet even with 100% scaling. The core difference is only 4.5%.

Perhaps it's include bug correction

[Shintai]

Perhaps it's include bug correction

Or was calculated on a Phenom with bug, hah

[JumpingJack]

Or was calculated on a Phenom with bug, hah

It is probably the way they calculated the 'mean', a geometric mean is not weighted to be necessarily linear with a linear dependent variable.

AMD is a huge fan of 'hiding' data behind geometric means. Most all of their benchmark presentations use this metholdology recently.

One reason I am not happy with AMD and have been so massively critical.

[nemrod]

The funny thing, is that the projected 9900 (2.6GHz) performs only 31% better than 9500 (2.2 GHz] when 9700 (2.4 GHz) performs 24% better...

So I vote for the shintai explanation, calculation done on a phenom

[JumpingJack]

The funny thing, is that the projected 9900 (2.6GHz) performs only 31% better than 9500 (2.2 GHz] when 9700 (2.4 GHz) performs 24% better...

So I vote for the shintai explanation, calculation done on a phenom

Doesn't make much sense does it.... Unless they are planning on some major bug fix.

[JumpingJack]

The 9500 vs 9600 is a..."miscalculation"...aswell. Its 2.2Ghz vs 2.3Ghz. They say 7% more performance. Yet even with 100% scaling. The core difference is only 4.5%.

Not to mention that it is contrary to all the 3rd part benches/data (see lead post of this thread )....