AMD 9950 BE vs Intel Q6600

D

dennisjai

2[H]4U
Joined
Jan 20, 2005
Messages
2,973
First off, those that have Frys don't chime in!

Okay, price for 9950BE (2.6ghz) is 180 vs Q6600 (2.4ghz) for 190

Anandtech had a review of the 9950BE right under the Q6600 or beating it in some applications/games..

Looks like AMD is competitive, i'm just wondering why everyone bashes on AMD saying they can't keep up. Sure Q6600 and 9950BE can hit 3.0ghz fairly easily, and I know Intel can overclock further.

Remember the AMD XP vs Intel P4 days?

I'm not looking to start a war, but both CPUs seems great and I'm torn between both of them.
 
You will get higher MIN (LOW) FPS with Phenom
Intel gives you higher average and max FPS
If you only run one application then intel is good
AMD is more of a multitasker and quick startups, if you want good performance running more than one application (that is working) amd is better.

Both are VERY FAST processors
 
dennisjai said:
First off, those that have Frys don't chime in!

Okay, price for 9950BE (2.6ghz) is 180 vs Q6600 (2.4ghz) for 190

Anandtech had a review of the 9950BE right under the Q6600 or beating it in some applications/games..

Looks like AMD is competitive, i'm just wondering why everyone bashes on AMD saying they can't keep up. Sure Q6600 and 9950BE can hit 3.0ghz fairly easily, and I know Intel can overclock further.

Remember the AMD XP vs Intel P4 days?

I'm not looking to start a war, but both CPUs seems great and I'm torn between both of them.
Click to expand...

Advantages of Q6600 over 9950BE:
- Better overclocking
- 10% Higher IPC
- Lower power consumption (95W vs 140W TDP)
- Comes with a stock HSF

I think thats worth a $10 premium...
 
9950BE clocks to about 3.5 GHz with 790GX
I have compared my computers (9750, 9500) with some Intel computers and total power consumption hasn't been higher comparing to Intel (65 nm). In fact it has been lower
 
harpoon said:
Advantages of Q6600 over 9950BE:
- Better overclocking
- 10% Higher IPC
- Lower power consumption (95W vs 140W TDP)
- Comes with a stock HSF

I think thats worth a $10 premium...
Click to expand...

I don't know about 9950BE but 9600BE comes with a better stock HSF that the Q6600.
 
dennisjai said:
First off, those that have Frys don't chime in!

Okay, price for 9950BE (2.6ghz) is 180 vs Q6600 (2.4ghz) for 190

Anandtech had a review of the 9950BE right under the Q6600 or beating it in some applications/games..
Click to expand...
The Intel CPUs used in the recent tests score lower than the ones they used in the Phenom 9850BE test.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3344&p=13
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3272&p=9

The difference may be caused by the use of an Intel G35 motherboard in the recent tests versus the X38 used previously.

Looks like AMD is competitive, i'm just wondering why everyone bashes on AMD saying they can't keep up. Sure Q6600 and 9950BE can hit 3.0ghz fairly easily, and I know Intel can overclock further
Click to expand...
The main reasons are probably because the Q6600 came out in February 2007 and is a low-end quad-core for Intel. Conversely, the Phenom 9950 was recently released, is the third clock upgrade to the Phenom lineup and is still slower than the Q6600 while using significantly more power. AMD is only competitive in the sense that their best quad-core is almost as good as Intel's worst quad-core.

Meanwhile, Intel has gone to 45nm products and has a 3GHz quad-core that uses about 65W.
 
Thanks for the constructive information SLee, I did forget that the 9950 just came out recently which is why amd phenoms were not competitive enough back then..

I guess the CPU i'll be waiting for is the 9850 on price drop =D (assuming its more than $15+ vs the 9950).
 
SLee said:
The main reasons are probably because the Q6600 came out in February 2007 and is a low-end quad-core for Intel. Conversely, the Phenom 9950 was recently released, is the third clock upgrade to the Phenom lineup and is still slower than the Q6600 while using significantly more power. AMD is only competitive in the sense that their best quad-core is almost as good as Intel's worst quad-core.
Click to expand...

And soon Intel will be out with i7 which is one year after phenom ;)
 
Atech said:
But one year ahead in performance...all that matters.
Click to expand...

That depends on what type of application you are running and most important, if you run more than one application. Phenom is designed for multithreading, intel has focus on singel threaded applications
 
kassler said:
That depends on what type of application you are running and most important, if you run more than one application. Phenom is designed for multithreading, intel has focus on singel threaded applications
Click to expand...

Oh yea, they added QPI and HyperThreading because they think single-CPU and single-threading is the future!
 
kassler said:
9950BE clocks to about 3.5 GHz with 790GX
I have compared my computers (9750, 9500) with some Intel computers and total power consumption hasn't been higher comparing to Intel (65 nm). In fact it has been lower
Click to expand...

3.5GHz would be a best case scenario, not all 9950BEs can hit 3.5GHz, in fact the vast majority can't, from what I've seen so far anyway. That said, even at 3.5GHz, a Phenom would be no match for the typical Q6600 OC @ 3.4 - 3.6GHz as it still trails by ~10% in IPC (and please don't start again on the Intel sucks at multithreading crap, we've heard it all before).

As for power consumption, the Phenom 9750 and 9500 are 95W chips, so power consumption should be similar to 65nm C2Qs which are also 95W. The 9950BE however is a 140W chip, although a 125W model is coming out soon.

kassler said:
I don't think Intel know themselves where hyperthreading will have advantages
Click to expand...
Nah, I think its just you that is confused. ;)
 
harpoon said:
Nah, I think its just you that is confused. ;)
Click to expand...
Maybe, but you need to program specifically and maybe compile the application for hyperthreading in order to use it effectively. If you don't then your application will be slower.
This type of development isn't common if you know what I mean.
 
Technically the code for Hyperthreading is handled by the BIOS it's designed to act as a separate thread or a virtual processor so there is no coding to be done on the application side. I would also consider the Q6600 to be a better buy. Most can easily reach 3GHz.
 
kassler said:
Maybe, but you need to program specifically and maybe compile the application for hyperthreading in order to use it effectively. If you don't then your application will be slower.
This type of development isn't common if you know what I mean.
Click to expand...

No, to a program/OS hyperthreading just appears as more cores. So as long as the OS schedules threads on the logical cores (which any multicore/multi-CPU OS is already capable of anyway), there's nothing specific you need to do as a programmer. This was demonstrated even in the early days with Pentium 4/Xeon, when things like 3dsmax would render quicker on HT systems.
And since Windows XP the OS thread scheduler is actually HT-aware, so it avoids potential performance pitfalls with HT.

In short, an application programmer doesn't need to do anything. Everything is already taken care of.
 
dennisjai said:
Looks like AMD is competitive, i'm just wondering why everyone bashes on AMD saying they can't keep up.
Click to expand...

Well here's one reason why:
http://www.legionhardware.com/document.php?id=770&p=0

In 3 out of 4 games (CoH, Sup Com, and UT3), it takes a X4 9950 at 3Ghz to match or at least get close to the performance of a stock Q6600 at 2.4Ghz. So imagine what the performance gap would be if the Q6600 was OC'd to 3Ghz+
 
Scali2 said:
In short, an application programmer doesn't need to do anything. Everything is already taken care of.
Click to expand...
And then you will have a slower application. You have to know how the processor works in order to gain speed.
 
kassler said:
And then you will have a slower application. You have to know how the processor works in order to gain speed.
Click to expand...

Why exactly would it be slower?
For each physical core, you get two logical cores.
Now, if I put an extra physical core on my CPU, do my applications get slower? I didn't think so.
So, if I put an extra logical core on my CPU, my applications won't get slower either. At the very least, the extra cores, physical or logical, are just ignored by the application (although they can be used by the OS to run other tasks in the background, but that goes beyond the scope of an application and its development).
 
Danny Bui said:
In 3 out of 4 games (CoH, Sup Com, and UT3), it takes a X4 9950 at 3Ghz to match or at least get close to the performance of a stock Q6600 at 2.4Ghz. So imagine what the performance gap would be if the Q6600 was OC'd to 3Ghz+
Click to expand...

Simple example counting FPS for one minute, just to show how FPS works in tests done from different sites

First situation - Resolution = 800x600
Computer 1 (Intel)
20 FPS for 10 seconds, 80 FPS for 40 seconds, 200 FPS for 10 seconds
(20 * 10 + 80 * 40 + 200 * 10) / 60 = 90

Computer 2 (AMD)
25 FPS for 10 seconds, 60 FPS for 40 seconds, 140 FPS for 10 seconds
(25 * 10 + 60 * 40 + 140 * 10) / 60 = 67,5

Second situation - Resolution = 1920x1200
The video card now cuts the FPS at 50 FPS

Computer 1
20 FPS for 10 seconds, 50 FPS for 40 seconds, 50 FPS for 10 seconds
(20 * 10 + 50 * 40 + 50 * 10) / 60 = 45

Computer 2
25 FPS for 10 seconds, 50 FPS for 40 seconds, 50 FPS for 10 seconds
(25 * 10 + 50 * 40 + 50 * 10) / 60 = ~46
 
kassler said:
Simple example just to show how FPS works in tests done from different sites

First situation - Resolution = 800x600
Computer 1 (Intel)
20 FPS for 10 seconds, 80 FPS for 40 seconds, 200 FPS for 10 seconds
(20 * 10 + 80 * 40 + 200 * 10) / 60 = 90

Computer 2 (AMD)
25 FPS for 10 seconds, 60 FPS for 40 seconds, 140 FPS for 10 seconds
(25 * 10 + 60 * 40 + 140 * 10) / 60 = 67,5

Second situation - Resolution = 1920x1200
The video card now cuts the FPS at 50 FPS

Computer 1
20 FPS for 10 seconds, 50 FPS for 40 seconds, 50 FPS for 10 seconds
(20 * 10 + 50 * 40 + 50 * 10) / 60 = 45

Computer 2
25 FPS for 10 seconds, 50 FPS for 40 seconds, 50 FPS for 10 seconds
(25 * 10 + 50 * 40 + 50 * 10) / 60 = ~46
Click to expand...

Yea, in Fantasyland such computers might actually exist too!
 
Scali2 said:
Why exactly would it be slower?
For each physical core, you get two logical cores.
Now, if I put an extra physical core on my CPU, do my applications get slower? I didn't think so.
Click to expand...
We where talking about hyperthreading, not more physical cores.
Two cores compared to one core with hyperthreading is two different areas
 
kassler said:
Simple example counting FPS for one minute, just to show how FPS works in tests done from different sites
Click to expand...

And? Do you have proof beyond a reasonable doubt that supports whatever you're arguing?

You know what? Lets both stop. I don't want this thread to end up like the last few threads where everyone proved you wrong about AMD CPUs. Unless you have some exciting new data to back up your hypothesis that AMD CPUs are better for gaming (which has been your claim in the last few threads) lets just stop.
 
kassler said:
We where talking about hyperthreading, not more physical cores.
Two cores compared to one core with hyperthreading is two different areas
Click to expand...

Yes, but the comparison is one core and one core with hyperthreading (or actually 4 cores and 4 cores with hyperthreading), since Intel is adding hyperthreading to its CPU, it's not replacing physical cores with logical ones, it adds logical cores to the physical ones. I was just saying that adding cores (regardless of whether they're physical or logical) doesn't make applications slower.
So again, why would that make applications run slower? What exactly would a programmer have to do?
 
kassler said:
Do you know anything about thread switching and thread synchronization?
Click to expand...

Ofcourse, I'm an MSc in Software Engineering, remember?
But I'm asking you to explain the point you're trying to make. You're just returning the question. You always do that when someone asks you to explain or prove something.

Why don't you explain first, and if I don't understand something, I can ask questions.
 
RamonGTP said:
You should ask yourself the very same question. You have been PROVEN wrong in EVERY thread you've posted in.
Click to expand...

Worst thing is, HyperThreading has been on the market for years. So his claims that Intel wouldn't know how it performs is rather silly.
Many of us have used and/or owned HyperThreading machines and are quite familiar with how they relate to non-HyperThreaded ones.

Aside from that, both Sun and IBM use similar technology in their server CPUs. Especially Sun's Niagara's CPUs are quite impressive. They have 8 cores with 8-way simulataneous multicore (so 64 logical cores).

But perhaps they just did that to make applications run slower?
 
It's unfortunate, but I don't think any amount of proof is going to do anything to this forum member. It's unfortunate too since newer members who don't yet know he hasn't got a clue may actually believe some of his BS if no one catches it in time.
 
kassler said:
That depends on what type of application you are running and most important, if you run more than one application. Phenom is designed for multithreading, intel has focus on singel threaded applications
Click to expand...

Stop doing drugs:
Intel Nehalem benchmarked: muscling to victory
Professional software, which is usually optimised for multi-core usage, shows Nehalem in an appropriately refulgent light; trampling on the C2E QX9770 in unceremonious fashion. Indeed, it's over 33 per cent faster than the aforementioned CPU in Pov-ray 64-bit, and over twice as potent as AMD's best. Scary, huh?
Click to expand...

Bottom line: Nehalem's very real performance numbers put incumbent Core 2 Extreme in the shade and pose AMD some really tough headaches.
Click to expand...
 
RamonGTP said:
It's unfortunate, but I don't think any amount of proof is going to do anything to this forum member. It's unfortunate too since newer members who don't yet know he hasn't got a clue may actually believe some of his BS if no one catches it in time.
Click to expand...

Since pretty much every single one of his post contains the same misinformation, I think there's a very simple solution to that problem...
 
Atech said:
Stop doing drugs:
Intel Nehalem benchmarked: muscling to victory
Click to expand...

Indeed, note also that Phenom ends last, at a distance, in every multithreading benchmark there.
So Intel doesn't even need Nehalem to beat AMD at multithreading, they're already far ahead (for those that didn't know yet... oh I guess that's just kassler).
Audio/video encoding, RAR compression, 3d rendering, gaming... you name it, Phenom is last in every single multithreading benchmark.
What's worse... they used the latest and greatest Phenom, only out for a few weeks... One of the Intels is the QX6800, of the first batch of quadcores launched by Intel, in early April 2007. AMD still can't catch it more than a year later.

But ofcourse you can ignore all this evidence, and still drone on about how AMD is better in multithreading in every bloody thread on this forum. I don't see what the point is, because nobody is going to buy into that nonsense. Even in the AMD section people are calling him out now :)
 
Scali2
You need to do more real programming. Its obvious that you just read about it and sits at forums and bashes AMD.
When you develop something you always have a reason why/how you do it. Scaling the code to more threads is also done for a reason (mostly for gaining speed). You don't create threads for fun..
If one core is running two threads it will always be slower compared to if the functionality was in one thread. Switching threads takes time, cache data etc will not be as effective, overhead for synchronization etc.
If you don't believe me then I would gladly prove it for you. You could try to do something in two threads, I do it in one and we will see who gets the fastest result if the code runs on the same PHYSICAL core.
 
kassler said:
Scali2
You need to do more real programming. Its obvious that you just read about it and sits at forums and bashes AMD.
Click to expand...

Heh, right. I've showed you where you can find examples of my work.
Now show me yours.

kassler said:
When you develop something you always have a reason why/how you do it. Scaling the code to more threads is also done for a reason (mostly for gaining speed). You don't create threads for fun..
Click to expand...

Now you're reversing the statement.
You claimed that applications got slower when HyperThreading was enabled.
That has nothing to do with scaling the code to more threads. The code is unchanged.

kassler said:
If one core is running two threads it will always be slower compared to if the functionality was in one thread. Switching threads takes time, cache data etc will not be as effective.
If you don't believe me then I would gladly prove it for you. You could try to do something in two threads, I do it in one and we will see who gets the fastest result if the code runs on the same PHYSICAL core.
Click to expand...

No need. Remember my multithreaded Marching Cubes implementation?
I already tested that on a Pentium 4 HT back in the day. It ran faster with two threads than with the original single-threaded code (iirc it was 160 fps vs 185 fps on that machine, a 2.8 GHz P4HT). Anyone can download and verify that for themselves.
The gain from the second thread is more than enough to cover the additional overhead. That's the whole idea behind HT obviously. Intel, IBM and Sun wouldn't bother if it only made software run slower.

I think you're in over your head here. You don't seem to understand how the execution pipeline in a modern processor works, and how a lot of units sit idle most of the time, because you cannot feed enough instructions to keep them busy in a single thread.
HT is just a way to get a more varied and less dependent tinstruction mix into the out-of-order buffer. Intel has some good architecture optimization manuals explaining the whole cycle of decoding, reordering, issuing and retiring instructions. I'd also like to direct you to Agner Fog's documentation on the subject.
Once you understand CPUs and code optimization at that level, the idea of HT is painfully obvious.
 
Scali2 said:
Heh, right. I've showed you where you can find examples of my work.
Now show me yours.
Click to expand...
It isn't possible and I don't think you would understand it. But one of the biggest CRM application is mostly done by me (two others also).
 
kassler said:
It isn't possible and I don't think you would understand it. But one of the biggest CRM application is mostly done by me (two others also).
Click to expand...

Nice how you ignore the rest. Guess you didn't understand it? :)
Here's something else I wrote, see if you understand this: http://www.megasecurity.org/Programming/optimize.html
Or this: http://verhoeven272.nl/cgi-bin/FSgz?fruttenboel/asm&Fruttenboel+assembly&asmtop&apj08&asmcontent

(X-Calibre was my 'full' nickname before it was shortened to Scali).
 
Scali2 said:
Nice how you ignore the rest. Guess you didn't understand it? :)
Click to expand...

yes I did but it takes to much time.
You know that is completely stupid to compare P4 with i7. P4 was a very bad processor and HT was a fix to solve some of the bad design that processor had.
 
kassler said:
yes I did but it takes to much time.
Click to expand...

No, you just don't know what you're talking about. You're not fooling anyone here.

kassler said:
You know that is completely stupid to compare P4 with i7. P4 was a very bad processor and HT was a fix to solve some of the bad design that processor had.
Click to expand...

What about the IBMs and Suns then?
And what about the fact that Hexus.net concludes that HT boosts performance in multithreaded scenarios on Nehalem?
"Last but not least, SMT (simultaneous multithreading) provides a healthy boost, too.
Looking back through the numbers, the 2.93GHz Nehalem naturally comes into its own when the cores, be they physical or virtual, are pushed by the software. When this happens, it's up to 33 per cent faster than a 3.2GHz Intel Core 2 Extreme QX9770, and some 50 per cent quicker than an equivalently-clocked (Kentsfield-based) Core 2 Quad CPU."

You're just lashing out at the P4 (which nobody cares about anymore... P4-bashing is so 1999), but you're not actually making any kind of argument for your point, despite the overwhelming evidence against it.
Really, why would anyone believe what you say? In fact, why would anyone even take you seriously at this point?
You're not really saying anything. You just make some unfounded claims, and never go into any kind of explanation. You just make personal attacks, or go off on a tangent about some irrelevant detail.
It's time to put up or shut up.
 
You must log in or register to reply here.
Back
Top