I wonder how many of these tests benefited from the new Intel speed boost technology the temporarily increase the CPU speed when the other cores not are heavily used, and hence, i wounder if these tests represent server load, that happens 24x7.
We Know Dunnington is 2 Billion Transistors, about 3 months ago, Nahalem was in hardlyspelledright, as 2-2.4 billion transistors units, lesser being lynnfield. Now above article states 800 million trans, this is typical of Bait & Switch in Industry. You buy 2 billion & recieve upon final investigation 800 million. Intel You Poo-Poo, at Times.drashek
Where are Linux drivers for NVIDIA chipsets buggy?
My AMD system with MCP61 chipset ran just fine out of the box, thank you (that is, after I custom built a kernel with workarounds for Barcelona's 298 and 309 errata).
And if you can't find the drivers, just look in your vanilla Linux kernel, not on the NVIDIA website. It's all there, ready to go.
Finally, although I very much like Nehalem's performance, my next system will probably still be an AMD one. After all, it's the bang for bucks that matters to me.
excuse me - it seems, sometimes, I still miss the difference between right side and left side:)...
The "old HT" way is, obviously, wrongly displayed in left not right side of the picture, as I previously stated..
I think the "The old is the new new" comment author is right about one thing - the new SMT is the same as old HT (HT was just a marketing term for SMT anyway). May be it was improved a bit, but it is the same old HT. It makes sense that SMT gives greater benefits to Nehalem than it did to P4, because Nehalem has a wider core, but other than that is about the same.
Right picture which should show us how the "old HT" worked shows us just how traditional Multithreading is done on non SMT CPU`s - by context switches which allows CPU to switch from one thread to another.
I think this excellent article explains this in simple and detailed way - http://arstechnica.com/old/content/2002/10/hyperthreading.ars.
Turbo mode is great for rigging benchmarks, you can claim a 3GHz Intel verses 3GHz Shanghi, run a single process benchmark which will run at 3.6GHz and beat the Shanghi easily.
Most Shanghis will let you overclock all four cores to 3.6GHz, ends up with lots of 2% either ways and draws. IF you could actually clock the Intel chip that fast...
Once AMD adds a turbo-boost kernel this bullshitting will move to some other rigged benchmarking setup.
Im sure you've all noticed it already. Sylvie is bashing NV lately, the rabbit we know to be pro-Isra...Intel, and Charlie, the paid-by-AMD anti-Nvidia/Intel propaganda machine, omg what's up with you Charlintel? You know what this means, right? The money people have turned Charliar against the only honest x86 mfg. of our times. Btw, Charlie is being too optimistic about Intel's new baby. Ever heard about a Quad-socket MCP3200 Opteron board, Charlie? Tyan and Supermicro have $700ish quad-opteron boards with Tri-SLI support. And with the latest cut in AMD pricing, you can build yourself a 16-core SLI workstation @ half the price of Intelahelum. I hope Nick Farrel comes in for the rescue xD
I just wanted to point out the old P4 HT was marketed working exactly like the new i7 HT, which makes sense since they are both OoO processors. What you are referring as the old P4 way of multithreading is the CMT way of multithreading used more often in in-order processors.
anandtech have put a nice review here:
http://it.anandtech.com/IT/showdoc.aspx?i=3536
nehalem beats the socks off everything non-nehalem. the only thing that beats it is dunnington quad socket, and that too by a small margin. and you can't really complain about that, because it's 24 core vs 8 cores of nehalem, and nehalem still comes pretty darn close.
90% of the world uses Windows whether you throw a temper tantrum or not. At least keep a Windows PC in the corner that can provide the content that everyone else can see. If I was so lame, I would be fired in a mouse click. Give me a break.
The article is a yawner. Nehalem server performance was already old news.
I wonder how many of these tests benefited from the new Intel speed boost technology the temporarily increase the CPU speed when the other cores not are heavily used, and hence, i wounder if these tests represent server load, that happens 24x7.
We Know Dunnington is 2 Billion Transistors, about 3 months ago, Nahalem was in hardlyspelledright, as 2-2.4 billion transistors units, lesser being lynnfield. Now above article states 800 million trans, this is typical of Bait & Switch in Industry. You buy 2 billion & recieve upon final investigation 800 million. Intel You Poo-Poo, at Times.drashek
My AMD system with MCP61 chipset ran just fine out of the box, thank you (that is, after I custom built a kernel with workarounds for Barcelona's 298 and 309 errata).
And if you can't find the drivers, just look in your vanilla Linux kernel, not on the NVIDIA website. It's all there, ready to go.
Finally, although I very much like Nehalem's performance, my next system will probably still be an AMD one. After all, it's the bang for bucks that matters to me.
excuse me - it seems, sometimes, I still miss the difference between right side and left side:)...
The "old HT" way is, obviously, wrongly displayed in left not right side of the picture, as I previously stated..
I think the "The old is the new new" comment author is right about one thing - the new SMT is the same as old HT (HT was just a marketing term for SMT anyway). May be it was improved a bit, but it is the same old HT. It makes sense that SMT gives greater benefits to Nehalem than it did to P4, because Nehalem has a wider core, but other than that is about the same.
Right picture which should show us how the "old HT" worked shows us just how traditional Multithreading is done on non SMT CPU`s - by context switches which allows CPU to switch from one thread to another.
I think this excellent article explains this in simple and detailed way - http://arstechnica.com/old/content/2002/10/hyperthreading.ars.
Turbo mode is great for rigging benchmarks, you can claim a 3GHz Intel verses 3GHz Shanghi, run a single process benchmark which will run at 3.6GHz and beat the Shanghi easily.
Most Shanghis will let you overclock all four cores to 3.6GHz, ends up with lots of 2% either ways and draws. IF you could actually clock the Intel chip that fast...
Once AMD adds a turbo-boost kernel this bullshitting will move to some other rigged benchmarking setup.
Im sure you've all noticed it already. Sylvie is bashing NV lately, the rabbit we know to be pro-Isra...Intel, and Charlie, the paid-by-AMD anti-Nvidia/Intel propaganda machine, omg what's up with you Charlintel? You know what this means, right? The money people have turned Charliar against the only honest x86 mfg. of our times. Btw, Charlie is being too optimistic about Intel's new baby. Ever heard about a Quad-socket MCP3200 Opteron board, Charlie? Tyan and Supermicro have $700ish quad-opteron boards with Tri-SLI support. And with the latest cut in AMD pricing, you can build yourself a 16-core SLI workstation @ half the price of Intelahelum. I hope Nick Farrel comes in for the rescue xD
I just wanted to point out the old P4 HT was marketed working exactly like the new i7 HT, which makes sense since they are both OoO processors. What you are referring as the old P4 way of multithreading is the CMT way of multithreading used more often in in-order processors.
anandtech have put a nice review here:
http://it.anandtech.com/IT/showdoc.aspx?i=3536
nehalem beats the socks off everything non-nehalem. the only thing that beats it is dunnington quad socket, and that too by a small margin. and you can't really complain about that, because it's 24 core vs 8 cores of nehalem, and nehalem still comes pretty darn close.
That would be a very grim day for us... consumers
90% of the world uses Windows whether you throw a temper tantrum or not. At least keep a Windows PC in the corner that can provide the content that everyone else can see. If I was so lame, I would be fired in a mouse click. Give me a break.
The article is a yawner. Nehalem server performance was already old news.
Another year of this and it's probably lights out for the Green Arrowed ones.