Can AMD salvage QFX with an in-house chipset?

Page 14 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
And soon I will be making this music with my new Turion X2 laptop.

Honest to god Baron I would love to hear this Gangsta Rap you're going to be cooking up on your new Turion laptop.

Although in music making, the Turion will get, how you say, 'smoked' by Core 2 systems.
 
Having looked at 14 pages of BS ranging from various suppositions about various hypothetical CPU's that may or not be made, vs. the cost of being married or not...

Can anyone point me to a cost - benefit analysis of why I should buy 10000 of these systems vs the 10000 systems currently used?

Bottom line. If I can write them off, and improve effeciency, then good. If its just an upgrade that does not improve effenciency, bad.


I'd have to say that teh biggest argument for this is apps that are high-bandwidth multi-threaded in nature. becasue of the inherent increeased power of two chips, this is not an SLI gamer's platform.

Quadro Sli could be used for those heavy duty jobs where the brute force method helps get things done faster. In those cases saving a month is worth more than saving $25-50/ month extra electricity bill.

Though most reviews were with pre-release boards, Anand manged to get temps and power down significantly in his tests (456W vs 713W load). I think the link with the higher power was using 8800 SLI.

Another case where this would come in handy is for developers who don't need gaming graphics. If for whatever reason neither C2Q nor Opteron suit your needs or preference, this thing will chew through C# compiling like a fiend.

Again because most of the time you are actually writing code (or setting rendering options), you won't hit anywhere near load 75% of the time ( maybe more, maybe less)


It's still as good an idea as dual socket anything else or quad core anything else. I think it wil shape up to be a Vista monster by Feb.

.............,-~*'`¯lllllll`*~,._
..........,-~*`lllllllllllllllllllllllllll¯`*-,
.....,-~*llllllllllllllllllllllllllllllllllllllllllll*-,
..,-*llllllllllllllllllllllllllllllllllllllllllllllllllllll.\
;*`lllllllllllllllllllllllllll,-~*~-,llllllllllllllllllll\
.\lllllllllllllllllllllllllll/.........\;;;;llllllllllll,-`~-,
..\lllllllllllllllllllll,-*...........`~-~-,...(.(¯`*,`,
..\llllllllllll,-~*.....................)_-\..*`*;..)
...\,-*`¯,*`)............,-~*`~................/.
....|/.../.../~,......-~*,-~*`;................/.\
..../.../.../.../..,-,..*~,.`*~*................*...\..
...|.../.../.../.*`...\...........................)....)¯`~,..
...|./.../..../.......)......,.)`*~-,............/....|..)...`~-,....
..././.../...,*`-,.....`-,...*`....,---......\..../...../..|.........¯```*~-,
...(..........)`*~-,....`*`.,-~*.,-*......|.../..../.../............\.........
.....*-,.......`*-,...`~,..``.,,,-*..........|.,*...,*...|..............\........
.........*,.........`-,...)-,..............,-*`...,-*....(`-,............\.......
............`-,.........`-,/...*-,___,,-~*....,-*......|...`-,..........\.......
............,~)...........)|............*,`*~--*........|......`-,.........\.....
 
How long does it take to compile code? I mean yikes, does it really take that long that you'd need four cores working on it?

A surprisingly long time actually, if it's a large project in C++. Compiling a C++ file at work on a 3GHz-ish P4 seems to take about as long as compiling a similarly-sized C file did on a Sparcstation 2 at 40MHz fifteen years ago.

Multiply that by maybe a thousand files in a decently-large project and you're talking about 40 minutes to do a full build. And I seem to remember reading a while back in an article about Microsoft that building Windows from scratch takes a day or more.

That said, I'm not sure whether that's due to CPU performance or disk performance: though most header files should be in the RAM cache after building a few files.

WIndows USED to take that long. It doesn't now that MS uses Opteron and X64. I beleive the last quote from BrianV was 6 hours.
 
/re-enter thread

I'd have to say that teh biggest argument for this is apps that are high-bandwidth multi-threaded in nature. becasue of the inherent increeased power of two chips, this is not an SLI gamer's platform.

Someone, preferably not Baron, correct me if I'm wrong but wasn't this platform marketed specifically at enthusiasts, the majority of which are gamers, and thus would want SLI?

That is your opinion. All you fanboys are out in force makin people's lives miserable becaue it does indeed close the gap at the high and leaves X6800 in the dust- especially in 64-bit Vista.

Please provide us with links of benchmarks where this "leaves X6800 in the dust"? Please highlight the 64-bit Vista benchmarks.

Were you not reading subjectively you would have read that that benchmark site was posted but it's on www.hwupgrade.com

@ Boduke --- you have been zapped by Baron nLogic®, don't worry the affects are short lasting.

Baron, why do you do this?? I pulled the 'smokes the x6800' benchmarks you linked and you are either incredibly dishonest or incredibly dumb.

The 'smoking the x6800' of course will occur if the application can split into more threads.... you failed to note that the FX-74 QFX is also soundly beaten, in the same 'smokin' ' benchmarks, by the QX6700. I will not regurgitate that data, but if take a look even in a couple of benchmarks the x6800 still manages to pull ahead (even when a NUMA OS is used -- BTW, XP is numa aware perhaps not well, but it is).



Before the QFX becomes a viable option over X6800 or other single socket solutions, more software than 3DS Max and a few encoders must become available. That is the only way to make it truly appealing to anyone but videographers and 3D artists. Even then, the QX6700 is a better option over the 4x4 for now. Maybe Barcelona will change that but based on the data you are not being honest when you continue to pound this drum.

Here is MS MSDN Dev (your favorite abbreviation) stating XP is NUMA aware:
http://msdn.microsoft.com/library/default.asp?url=/library/en-us/DevTest_g/hh/DevTest_g/BootIni_17b4305e-23ac-40ea-99db-4858b29a5d66.xml.asp

NUMA is supported in all editions of Windows XP and Windows Vista, and in the Enterprise and Datacenter editions of Windows Server 2003.

This is not to say Vista will not do it better, but the NUMA hit will always be there, no matter how hard you try to avoid it.

Both Extremesys and Hexus knew of this and enabled it.

You continue to spread FUD.

Thanks Jack. I didn't follow his link because I knew I'd find exactly what you posted. I exited the thread earlier because of said Baron nLogic®. When I saw his postings I had to comment, I obviously I was sucked in. Fried a synapse with that one I think. :)
 
/re-enter thread

I'd have to say that teh biggest argument for this is apps that are high-bandwidth multi-threaded in nature. becasue of the inherent increeased power of two chips, this is not an SLI gamer's platform.

Someone, preferably not Baron, correct me if I'm wrong but wasn't this platform marketed specifically at enthusiasts, the majority of which are gamers, and thus would want SLI?

That is your opinion. All you fanboys are out in force makin people's lives miserable becaue it does indeed close the gap at the high and leaves X6800 in the dust- especially in 64-bit Vista.

Please provide us with links of benchmarks where this "leaves X6800 in the dust"? Please highlight the 64-bit Vista benchmarks.

Were you not reading subjectively you would have read that that benchmark site was posted but it's on www.hwupgrade.com

@ Boduke --- you have been zapped by Baron nLogic®, don't worry the affects are short lasting.

Baron, why do you do this?? I pulled the 'smokes the x6800' benchmarks you linked and you are either incredibly dishonest or incredibly dumb.

The 'smoking the x6800' of course will occur if the application can split into more threads.... you failed to note that the FX-74 QFX is also soundly beaten, in the same 'smokin' ' benchmarks, by the QX6700. I will not regurgitate that data, but if take a look even in a couple of benchmarks the x6800 still manages to pull ahead (even when a NUMA OS is used -- BTW, XP is numa aware perhaps not well, but it is).



Before the QFX becomes a viable option over X6800 or other single socket solutions, more software than 3DS Max and a few encoders must become available. That is the only way to make it truly appealing to anyone but videographers and 3D artists. Even then, the QX6700 is a better option over the 4x4 for now. Maybe Barcelona will change that but based on the data you are not being honest when you continue to pound this drum.

Here is MS MSDN Dev (your favorite abbreviation) stating XP is NUMA aware:
http://msdn.microsoft.com/library/default.asp?url=/library/en-us/DevTest_g/hh/DevTest_g/BootIni_17b4305e-23ac-40ea-99db-4858b29a5d66.xml.asp

NUMA is supported in all editions of Windows XP and Windows Vista, and in the Enterprise and Datacenter editions of Windows Server 2003.

This is not to say Vista will not do it better, but the NUMA hit will always be there, no matter how hard you try to avoid it.

Both Extremesys and Hexus knew of this and enabled it.

You continue to spread FUD.


ALl of the benches from Vista showed that only one or two things don't show the QFX right up with the C2Q. SInce the price is the same at $1000, you can compare clock speed and price.

I don't want to be associated with you so Intel can thank you for me pushing AMD.

I said simply that I USE MULTITHREADED APPS THAT RUN UNDER 64bit VISTA. X6800 DOESN'T HAVE A CHANCE. If I get playable frame rates, I'm happy. I don't need the highest fps.

In your minds if you get the most fps (above what's actually perceptible) you are superior in your thoughts, but that's BS, as you like to say.
I will not try to dissuade you from your purchases. If you do the same these kinds of back and forth things won't happen.

I want QFX as an upgrade to my current system.

END OF STATEMENT.

I will be whatever you want as long as it's MY CHOICE. WRONG OR RIGHT. Though PC choices don't enter into the realm of EARTH-SHATTERING.
 
ALl of the benches from Vista showed that only one or two things don't show the QFX right up with the C2Q. SInce the price is the same at $1000, you can compare clock speed and price.

Now I'm just answering to get my post count up...BUT...

$1000 QFX = 3Ghz
$1000 C2Q = 2.66Ghz

Not a clock/clock comparison. But maybe I'm missing something.

For the same cost we're looking at a processor pair @3Ghz that is "right up there" with a single package @2.66 Ghz (tho still a pair granted AND using an "antiquated" FSB) ...

Hmmmm......
 
I don't want to be associated with you so Intel can thank you for me pushing AMD.
What a crock of sh!t!! You were jerkin' off to Hector Ruiz' picture long before Jack became a forum member.If Jack is responsible for you pushing AMD, you are responsible for 100's of AMD lovers dissociating themselves from AMD. Your arrogance, and naivity never cease to amaze. :roll:
 
I don't want to be associated with you so Intel can thank you for me pushing AMD.
What a crock of sh!t!! You were jerkin' off to Hector Ruiz' picture long before Jack became a forum member.If Jack is responsible for you pushing AMD, you are responsible for 100's of AMD lovers dissociating themselves from AMD. Your arrogance, and naivity never cease to amaze. :roll:

Just direct them all downwind with the rest.
 
ALl of the benches from Vista showed that only one or two things don't show the QFX right up with the C2Q. SInce the price is the same at $1000, you can compare clock speed and price.

Now I'm just answering to get my post count up...BUT...

$1000 QFX = 3Ghz
$1000 C2Q = 2.66Ghz

Not a clock/clock comparison. But maybe I'm missing something.

For the same cost we're looking at a processor pair @3Ghz that is "right up there" with a single package @2.66 Ghz (tho still a pair granted AND using an "antiquated" FSB) ...

Hmmmm......


I just saw a Voyager 16GB USB drive and I thought to myself SuperFetch, perhaps the cure for your NUMA woes on Vista x64.

A 16GB drive would hold the RAM for ANY game 2 or 3 times over, so obviously swapping is a factor without NUMA. If NUMA works well with it (hmmm, don't I know some of the kernel developers?), then a single threaded app will never need to cross socket boundaries.
 
I just saw a Voyager 16GB USB drive and I thought to myself SuperFetch, perhaps the cure for your NUMA woes on Vista x64.

A 16GB drive would hold the RAM for ANY game 2 or 3 times over, so obviously swapping is a factor without NUMA. If NUMA works well with it (hmmm, don't I know some of the kernel developers?), then a single threaded app will never need to cross socket boundaries.

What a LOAD OF CRAP!

A 16GB USB stick ain't going to do ANYTHING to NUMA! Hell, I wouldn't put my swapfile on those things. What'd happen if you pulled it out? The OS asks for memory stored in the swapfile, which is gone... take a wild guess what would happen. (btw, I keep my swap file on a 15K SCSI hard drive)

1. USB latency is SKY HIGH compared to SYSTEM RAM!!
2. USB devices are not allocated memory space in a system!!!!
3. USB devices go through a host controller, and a driver! This adds latency!
4. Flash memory is SLOW AS HELL!! 10 MB/sec is PRETTY DAMNED AWESOME for a flash stick!
 
I just saw a Voyager 16GB USB drive and I thought to myself SuperFetch, perhaps the cure for your NUMA woes on Vista x64.

A 16GB drive would hold the RAM for ANY game 2 or 3 times over, so obviously swapping is a factor without NUMA. If NUMA works well with it (hmmm, don't I know some of the kernel developers?), then a single threaded app will never need to cross socket boundaries.

What a LOAD OF CRAP!

A 16GB USB stick ain't going to do ANYTHING to NUMA! Hell, I wouldn't put my swapfile on those things. What'd happen if you pulled it out? The OS asks for memory stored in the swapfile, which is gone... take a wild guess what would happen. (btw, I keep my swap file on a 15K SCSI hard drive)

1. USB latency is SKY HIGH compared to SYSTEM RAM!!
2. USB devices are not allocated memory space in a system!!!!
3. USB devices go through a host controller, and a driver! This adds latency!
4. Flash memory is SLOW AS HELL!! 10 MB/sec is PRETTY DAMNED AWESOME for a flash stick!


Why does it have to be a load of crap? If you don't believe that MS knows what they're doing that's one thing but SuperFetch is designed to cache certain things. It just adds another source for moving NON-PRIORITY thread data.

This of course doesn't mean that games or anything else run from it. It means that it is another way of maintaining consistency for dual socket systems that use singlethreaded apps.

USB 2.0 is 480mbps, 60MB/s and Flash runs at up to 16.6MB/s. We're only talking about swapping out RAM so that a game can take 2GB from one socket. Since people can run BF2 with ONLY 2GB, obviously the swapping mechanism works well, it just needs refinement to maintain all data in one socket when the app uses 2 or less cores.

It's funny why AMD is not supporting 4GB per socket because then it would hardly be an issue.

That would have to be my only "issue" with it. 8GB wouldn't encroach on server/wksta since any DDR2 board can hold 8GB.
 
ALl of the benches from Vista showed that only one or two things don't show the QFX right up with the C2Q. SInce the price is the same at $1000, you can compare clock speed and price.

Now I'm just answering to get my post count up...BUT...

$1000 QFX = 3Ghz
$1000 C2Q = 2.66Ghz

Not a clock/clock comparison. But maybe I'm missing something.

For the same cost we're looking at a processor pair @3Ghz that is "right up there" with a single package @2.66 Ghz (tho still a pair granted AND using an "antiquated" FSB) ...

Hmmmm......


I just saw a Voyager 16GB USB drive and I thought to myself SuperFetch, perhaps the cure for your NUMA woes on Vista x64.

A 16GB drive would hold the RAM for ANY game 2 or 3 times over, so obviously swapping is a factor without NUMA. If NUMA works well with it (hmmm, don't I know some of the kernel developers?), then a single threaded app will never need to cross socket boundaries.

Actually my cure is to turn superfetch off. It's garbage. And what does what you just posted have to do with my pointing out that it is not a clock/clock comparison?
 
Actually my cure is to turn superfetch off. It's garbage. And what does what you just posted have to do with my pointing out that it is not a clock/clock comparison?

My statement was that because the prices of FX74 and C2Q are the same, you can look at FX70 for clock comparison and it of course is slower than the FX74, so by definition it's slower than C2Q.

Tom's own CPU guide takes price as a factor so price wise, you can compare the higher clock. Either Intel would have to drop their price to $599 or AMD would need to raise the price of FX70. Neither seems likely, but you can rest assured that I am comparing to FX70 because I don't think $999 is in the cards even for two CPUs.

Because clock for clock C2 is faster, no one should expect QFX to make AMD faster, but there are psychos like me who love the idea and wish they had done it two years ago.

Sure, I wish they released 65nm shrinks, but other than that I only see a HUGE upgrade from my 4400+ for what I buy a PC for, and it's not games, for the 1000000th time. Gamers would be fools to even buy C2Q.

A 6700 with 8800GTX comes in at just the price of the C2Q. But as to the point of the post, I think that Vista and a more refined chipset - even from nVidia - will improve that gaming perf to just equal or slightly better.

Multi-threaded apps are already seeing the scaling I expected.
 
Well, the QFX has been released and while certain scenarios show incredible promise, certain areas are also saddled with too much baggage.

The fact that the Opteron dual can be outfitted with SLI in a wksta and offer good perf without these power levels implies that the total package could have been done better.

The Opteron 285 runs at 2.6GHz and this graph shows that without the additional SLI power AMD runs at 322W and the dual 5160 runs at 267W (full load).

Looking at the varous articles around teh web, it seems as though only Anand managed to actually find suitable tasks that were reasonable for multi-tasking. In his case he used BluRay movies which totally killed all the dual core systems.

His power numbers were also at least 100W lower than other test sites. He turned on CnQ and got the idle temps down to within 4W of the C2Q system. Of course this didn't dent the 456W the system drew at "full load" with an 8800GTX ( which I believe draws 225W+ ).

And these are FX74 numbers. FX70 is shown to use even less so I believe OEMs can get reasonable wksta power levels out of it in teh next few months, especially if AMD releases a new rev( they sorely need to drop power by at least 10%- perhaps more and the lessons learned can help get Agena down below the reported 125W)

Hexus is also reporting that they can show a defect in the NUMA implementation of the Asus board BIOS ( this post was going to be called "Did Asus and nVidia drop the QFX ball") that maybe why games are suffering so much from latency problems.

One review ( most are posted at AMDZone) stated that AMD is reporting that the Interleave mode will need to be turned off for Vista and on for XP.

I believe AMD reported that they would release their own branded chipset and hopefully it will be less power hungry than the 680a, which is reported to use more power than even 975X. Having two of them surely doesn't help. nVidia does have a two socket SLI hipset in the 3600 and ASUS' implementation is only $300. Even the $400 Asus 680i for Intel implements less PCIe for less power reserves.

Because 7950GT and the probably forthcoming 8950GT only require two slots for Quad SLI. I can see the need for 4 low end GPUs for certain content creators but even 3 PCIe slots can't really be used right now as no "Havok" type apps or cards have been released, except for the server (AMD Stream).

Only time will tell if AMD had planned to create an entire reference system based on an Ati chipset while allowing nVidia to be the launch partner.

But the real judgement is that only expert builders will make QFX something not too loud or hot, while Vista X64 may do wonders for it in multithreaded apps so it is not yet ready for prime time.

Let's go AMD! Show your true potential.

No.

PWNED!!

-The String
 
My purchases are NOT based on the competition, they are based on my preferred CPU vendor having what I want, which int his case is two sockets.


Baron, Baron, Baron

WHAT are you doing? This thread came back from the brink of nuclear holocost flamage so well, and we were all playing so nice, and then you make this statement.

How many times have you said you would never by from Intel?

Since I have never seen you refer to Cyrix, TI, or any of the other itty bitty proccessor manufacturers, and since we are talking x86 class, that leaves 2 choices...Intel and AMD.
You have definatively declared, many times, you would never buy from Intel. Leaving only AMD. Meaning no choice. Meaning no competition. Meaning any other reasons you give are completely and totally pointless/moot/passe'/discutable/strittig etc etc etc and valuless.
Meaning your "prefered" vendor has what you want, regardless of what it is since its "what they got or nothin' "

God, what were you thinking when you wrote this. Stop it. Play nice/right
 
:) :) Well, to be honest, I haven't been too nice --- but frankly he crapped his own thread.... he asks a question, doesn't like the answer then pulls a 'thumb' from his butt.


:O

Jack

Butt (pun intended) thats the only time threads degrade, when a few select people's thumbs and other various items come out of their butts. And of those, 3 of 4 have been banned, with 1 converted. :wink:
 
:) :) Well, to be honest, I haven't been too nice --- but frankly he crapped his own thread.... he asks a question, doesn't like the answer then pulls a 'thumb' from his butt.


:O

Jack

Perhaps the tone of the answers lent little credence to any "supposed" argument correct or not.

Some reported that it was OK to buy Opteron but not QFX. That pretty much sealed the deal. Remember that MS has not said that RC2 of Vista was the end of tweaks to the consumer versions.

Memory handling (profiling could help) is the only obstacle to single threaded perf achieving the benefits of the extra channels and bandwidth.

Profile lookups would enable the OS to accurately swap out data from threads that lower in priority as a new app loads. NUMA Problem solved.

Obviously MS thinks Flash is or will be fast enough to cache static data. A 16GB drive would cache a helluva lot of apps. That means that things can be flushed to disk as large memory footprint apps (games are really the only single/dual threaded examples) are loaded to a socket's RAM banks.

But it always the same. It's faster than most and slower than some.
 
:) :) Well, to be honest, I haven't been too nice --- but frankly he crapped his own thread.... he asks a question, doesn't like the answer then pulls a 'thumb' from his butt.


:O

Jack

Perhaps the tone of the answers lent little credence to any "supposed" argument correct or not.

Some reported that it was OK to buy Opteron but not QFX. That pretty much sealed the deal. Remember that MS has not said that RC2 of Vista was the end of tweaks to the consumer versions.

Memory handling (profiling could help) is the only obstacle to single threaded perf achieving the benefits of the extra channels and bandwidth.

Profile lookups would enable the OS to accurately swap out data from threads that lower in priority as a new app loads. NUMA Problem solved.

Obviously MS thinks Flash is or will be fast enough to cache static data. A 16GB drive would cache a helluva lot of apps. That means that things can be flushed to disk as large memory footprint apps (games are really the only single/dual threaded examples) are loaded to a socket's RAM banks.

But it always the same. It's faster than most and slower than some.

Baron,

Cmon now. Be honest...and realistic. A flash drive the cure for NUMA?
Please, spare me. "MS is proud to announce the US release of Vista, for a limited time only with a FREE 16GB flash drive to make it work right. We appologize for tying up one of your USB ports instead of listening to our engineers over our PR department."


How about this solution. Let all programs load a few seconds slower instead of making a few load marginally faster while at same time reducing the footprint of the bloatbeast and improving overall performance by letting the OS run the apps instead of wasting resourses trying to second guess what the user is going to do next.
 
I find it appalling that a 20MB/s 100-microsecond link is advocated as a remedy for design deficiencies in a 4GB/s 100-nanosecond interconnect. Or that swap file latency could be a noticeable performance factor when you have twice as much RAM as you need to perform your most intensive task.

The supposed requirements for getting 4x4 to perform are a bit ridiculous. With current market prices for DRAM and the fact that cheaper memory is one of the marketing points of 4x4, inefficient use of memory banks becomes a huge drawback.
 
God, I hate Microsoft.

I hate having to dig through the dllcache deleting files to prevent movie maker, msmessenger and Netmeeting from "fixing" themselves courtesy of the Windows File Protection service. I hate the Windows File Protection service. I hate wasting time trying to figure out how to delete the now empty yet ferociously protected file file folders for the usless MS Bloatware I deleted.

I hate it when people make excuses for MS bloatware...before its even released.

Opinion: Windows is like the Tasmanian Devil. Ugly, mean and pointless in the greater scheme of things