Overclock.net › Forums › Industry News › Hardware News › [TechReport] Intel gives peek into Nehalem bag of tricks
New Posts  All Forums:Forum Nav:

[TechReport] Intel gives peek into Nehalem bag of tricks - Page 3

post #21 of 55
Quote:
Originally Posted by pauldovi View Post
Nehalem is going to be a beast.
Couldn't have said it better myself.
Lee XT
(17 items)
 
  
CPUMotherboardGraphicsRAM
AMD FX-6300 Asus M5A97 SAPPHIRE Radeon HD 7850 AMD 4GB DDR3 1333MHZ 
RAMRAMRAMHard Drive
AMD 4GB DDR3 1333MHZ AMD 4GB DDR3 1333MHZ AMD 4GB DDR3 1333MHZ OCZ Vertex 4 256GB 
CoolingOSMonitorKeyboard
Corsair H80 Windows 8.1 Pro MCE Dell P2414H WHXV7  Microsoft Generic 
PowerCaseMouseMouse Pad
Ultra 600W Limited Edition NZXT Black Steel Razer Deathadder Razer Goliath 
Audio
Realtek HD Audio 
  hide details  
Reply
Lee XT
(17 items)
 
  
CPUMotherboardGraphicsRAM
AMD FX-6300 Asus M5A97 SAPPHIRE Radeon HD 7850 AMD 4GB DDR3 1333MHZ 
RAMRAMRAMHard Drive
AMD 4GB DDR3 1333MHZ AMD 4GB DDR3 1333MHZ AMD 4GB DDR3 1333MHZ OCZ Vertex 4 256GB 
CoolingOSMonitorKeyboard
Corsair H80 Windows 8.1 Pro MCE Dell P2414H WHXV7  Microsoft Generic 
PowerCaseMouseMouse Pad
Ultra 600W Limited Edition NZXT Black Steel Razer Deathadder Razer Goliath 
Audio
Realtek HD Audio 
  hide details  
Reply
post #22 of 55
Quote:
Originally Posted by darkcloud89 View Post
I wouldn't be so sure about this. I commonly see that people think that AMD has no answer to Nehalem until Bulldozer. Really this isn't correct because Nehalem is actually Intel's answer to Barcelona. How do I make this conclusion? Barcelona and Nehalem are pretty similar chips.

AMD's transition from K8 to Barcelona to K10.5/Shanghai is very similar to Intel's transition from Netburst to Conroe to Nehalem. Conroe brought improved single-threaded performance and IPC (as well as efficiency) compared to Netburst, and the integration of the memory controller should give Nehalem scalability. AMD, on the other hand, released Barcelona and took care of the scalability improvements first by improving cHT links and a monolithic quad core design. And if they stay on track, Shanghai should bring improved IPC, power requirements/thermals (transition to HKMG 45nm), and single threaded performance. I guess you could say both companies were headed towards the same destination, but took different paths to get there.

If AMD can deliver the IPC improvements with the transition to 45nm in Shanghai, then that should suffice as Nehalem's competitor. Not only would K10.5 be faster per clock, but HKMG should allow it to clock much higher than it does now. Phenom was a ridiculously complex die to manufacture at 65nm and I think the process is holding it back more than most people think, and the 45nm transition if executed timely may open some eyes. The performance advantage with Nehalem will not be as big with single-threaded performance as it will be in multithreaded. AMD was already very good when it came to multithreaded performance, so really they just need to improve single-threaded.

EDIT: I see my post from the other thread was already quoted
The problem is that Netburst to Core2 to Nehalm is 3 generations while K8 to k10 to k10.5 is 2.5. I very cautious about Shanghai and its IPC improvements. Im just hoping AMD can deliver.
OT: When is AMD getting SSE4?
Edited by Somenoob - 6/17/08 at 8:20pm
Lenovo W520
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2630QM idk lol Quadro 2000M 8 Gigabytes 
Hard DriveOSMonitorCase
500 7200RPM Windows 8 Pro + Ubuntu 15.6 1920x1080 Lenovo plastic 
  hide details  
Reply
Lenovo W520
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2630QM idk lol Quadro 2000M 8 Gigabytes 
Hard DriveOSMonitorCase
500 7200RPM Windows 8 Pro + Ubuntu 15.6 1920x1080 Lenovo plastic 
  hide details  
Reply
post #23 of 55
Oh wow, just as I get a 775 CPU something new comes up.
    
CPUMotherboardGraphicsRAM
E8400 GIGABYTE GA-P35-DS3L eVGA 9800GT Stock Speeds GSkill 4GB (2x2GB) PC6400 
Hard DriveOptical DriveOSMonitor
Samsung F3 500GB Samsung 22X DVD Drive Windows 7 64-bit 22" Acer x223w 
KeyboardPowerCaseMouse
IBM Model M Model 1391401 Antec EarthWatts 500W Troglodytic Tech Terminal Razer Deathadder 
Mouse Pad
Steelseries Qck+ 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
E8400 GIGABYTE GA-P35-DS3L eVGA 9800GT Stock Speeds GSkill 4GB (2x2GB) PC6400 
Hard DriveOptical DriveOSMonitor
Samsung F3 500GB Samsung 22X DVD Drive Windows 7 64-bit 22" Acer x223w 
KeyboardPowerCaseMouse
IBM Model M Model 1391401 Antec EarthWatts 500W Troglodytic Tech Terminal Razer Deathadder 
Mouse Pad
Steelseries Qck+ 
  hide details  
Reply
post #24 of 55
Quote:
Originally Posted by darkcloud89 View Post
I wouldn't be so sure about this. I commonly see that people think that AMD has no answer to Nehalem until Bulldozer. Really this isn't correct because Nehalem is actually Intel's answer to Barcelona. How do I make this conclusion? Barcelona and Nehalem are pretty similar chips.

AMD's transition from K8 to Barcelona to K10.5/Shanghai is very similar to Intel's transition from Netburst to Conroe to Nehalem. Conroe brought improved single-threaded performance and IPC (as well as efficiency) compared to Netburst, and the integration of the memory controller should give Nehalem scalability. AMD, on the other hand, released Barcelona and took care of the scalability improvements first by improving cHT links and a monolithic quad core design. And if they stay on track, Shanghai should bring improved IPC, power requirements/thermals (transition to HKMG 45nm), and single threaded performance. I guess you could say both companies were headed towards the same destination, but took different paths to get there.

If AMD can deliver the IPC improvements with the transition to 45nm in Shanghai, then that should suffice as Nehalem's competitor. Not only would K10.5 be faster per clock, but HKMG should allow it to clock much higher than it does now. Phenom was a ridiculously complex die to manufacture at 65nm and I think the process is holding it back more than most people think, and the 45nm transition if executed timely may open some eyes. The performance advantage with Nehalem will not be as big with single-threaded performance as it will be in multithreaded. AMD was already very good when it came to multithreaded performance, so really they just need to improve single-threaded.

EDIT: I see my post from the other thread was already quoted
Just because you can find abstract similiarities between 2 processors doesn't make them related. Intel didn't need to answer K10, and if it did, Penryn is the closest you will get.

When you lead the pack, you don't "answer" to your competitors, they answer to you.

Just because the siginifance of Penryn to Nehalem is similar to K8 to K10, doesn't make them at all related on a competitive basis. The logic just isn't there.

So you think all AMD needs is a process revision and IPC improvements to compete with a new Intel macroarchitecture? Realize that AMD not only has to compete on a performance level with Intel but also a manufacturing one. AMD may be on 45nm, but they are still using 200mm wafers whereas Intel has the 300mm wafers.

Intel is introducing a whole lot more than simply IPC improvements. Early benchmarks are showing 40% inprovment over Core 2 with launch clocks up to 3.2Ghz. Can AMD compete with those clocks at that performance?
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
post #25 of 55
Quote:
Originally Posted by Somenoob View Post
The problem is that Netburst to Core2 to Nehalm is 3 generations while K8 to k10 to k10.5 is 2.5. I very cautious about Shanghai and its IPC improvements. Im just hoping AMD can deliver.
OT: When is AMD getting SSE4?
Current Phenoms have SSE4a support, which is AMD's version of SSE4.

I'm being very level-headed with Shanghai as well. I think that if they can improve single-threaded performance by about 20-30% it will be a very capable competitor to Intel on the desktop front. If they do improve per-clock performance, also remember that HKMG 45nm is going to allow them to clock the CPUs much higher than they are now, it's possible that 3GHz will be the low end.
post #26 of 55
Quote:
Originally Posted by darkcloud89 View Post
Current Phenoms have SSE4a support, which is AMD's version of SSE4.

I'm being very level-headed with Shanghai as well. I think that if they can improve single-threaded performance by about 20-30% it will be a very capable competitor to Intel on the desktop front. If they do improve per-clock performance, also remember that HKMG 45nm is going to allow them to clock the CPUs much higher than they are now, it's possible that 3GHz will be the low end.
And these numbers (20-30% and 3Ghz) is coming from where? Intel has already demoed their performance, so we know that is concrete. Just like Phenom/Barcelona, Shanghai is all talk. Will it flop like it too?
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
post #27 of 55
Quote:
Originally Posted by pauldovi View Post
Just because you can find abstract similiarities between 2 processors doesn't make them related. Intel didn't need to answer K10, and if it did, Penryn is the closest you will get.

When you lead the pack, you don't "answer" to your competitors, they answer to you.

Just because the siginifance of Penryn to Nehalem is similar to K8 to K10, doesn't make them at all related on a competitive basis. The logic just isn't there.

So you think all AMD needs is a process revision and IPC improvements to compete with a new Intel macroarchitecture? Realize that AMD not only has to compete on a performance level with Intel but also a manufacturing one. AMD may be on 45nm, but they are still using 200mm wafers whereas Intel has the 300mm wafers.

Intel is introducing a whole lot more than simply IPC improvements. Early benchmarks are showing 40% inprovment over Core 2 with launch clocks up to 3.2Ghz. Can AMD compete with those clocks at that performance?
You hit the nail on the head there. It is true that intel is introducing technology that AMD has used for a while, but this does not mean that it is attempting to answer AMD releases. With the transition to C2D, the FSB interface is seriously starting to hold Intel back. QPI is a natural evolution to the Conroe architecture, as is a native quad design.
It goes to eleven
(13 items)
 
  
CPUMotherboardGraphicsRAM
E6300 DS3 EVGA 8600GTS 2GB XMS2 DDR2-800 
Hard DriveOSMonitorKeyboard
1.294 TB Arch Linux/XP Samsung 226bw Eclipse II 
PowerCaseMouse
Corsair 520HX Lian-Li v1000B Plus G7 
  hide details  
Reply
It goes to eleven
(13 items)
 
  
CPUMotherboardGraphicsRAM
E6300 DS3 EVGA 8600GTS 2GB XMS2 DDR2-800 
Hard DriveOSMonitorKeyboard
1.294 TB Arch Linux/XP Samsung 226bw Eclipse II 
PowerCaseMouse
Corsair 520HX Lian-Li v1000B Plus G7 
  hide details  
Reply
post #28 of 55
Quote:
Originally Posted by pauldovi View Post
Just because you can find abstract similiarities between 2 processors doesn't make them related. Intel didn't need to answer K10, and if it did, Penryn is the closest you will get.

When you lead the pack, you don't "answer" to your competitors, they answer to you.
Really? Nehalem is by all means meant to be a server processor. The improvements made over Conroe/Penryn were meant to improve multithreaded performance and the IMC was to allow it to scale better beyond 1 and 2 socket systems. The Opteron by far outperforms Xeons in servers with more sockets. Intel is not the leader here. Barcelona was created to compete in this space, and Nehalem is the answer to Barcelona. Their respective architectures on desktop machines are irrelevant, it means nothing in comparison to server loads.

Quote:
So you think all AMD needs is a process revision and IPC improvements to compete with a new Intel macroarchitecture? Realize that AMD not only has to compete on a performance level with Intel but also a manufacturing one. AMD may be on 45nm, but they are still using 200mm wafers whereas Intel has the 300mm wafers.
First of all, I already stated that AMD's manufacturing capacity is inferior. That is why it would have been stupid of them to counter Core2 on the desktop because Intel would still be manufacturing more and could sell them cheaper. This has nothing to do with the relative performance or significance of the underlying architecture.

I think a process revision will help AMD a lot more than you might realize. Designing a monolithic quad core on a 65nm process may have not been the best idea, it really is too big. Clock speed is arguably Phenom's biggest problem right now due to thermal management and power requirements. A die shrink to 45nm will help this and allow it to clock higher. HKMG will further alleviate the problem allowing it to clock even higher. This should come right around the same time as the 20% IPC improvement with Shanghai. I'm pretty confident that a 20% per-clock improvement as well as a dramatic clock speed increase would put Shanghai higher in single-threaded performance than Penryn. Per-clock Phenom almost matches Penryn already, any IPC improvement should take it farther.

Quote:
Intel is introducing a whole lot more than simply IPC improvements. Early benchmarks are showing 40% inprovment over Core 2 with launch clocks up to 3.2Ghz. Can AMD compete with those clocks at that performance?
It isn't a matter of whether or not it's faster than Core2, but in what context is it faster. The improvements in Nehalem were mostly all done with regard to scaling and multithreading. Yes the benchmarks did show 40% improvement compared to Core2, but in what context was this? That's right, it was multithreading. What happened when Nehalem was compared to Penryn using only a single thread?



Quote:
Originally Posted by pauldovi View Post
And these numbers (20-30% and 3Ghz) is coming from where? Intel has already demoed their performance, so we know that is concrete. Just like Phenom/Barcelona, Shanghai is all talk. Will it flop like it too?
They are coming from what AMD is claiming. Notice that I kept saying if they can get the improvements in. I obviously have no idea whether or not the K10.5 revision will make it into the 45nm process or not. AMD did claim that Barcelona would be faster, and yes desktop users were disappointed with its performance. But the reason is because they took "faster" out of context. It was meant to be faster in multithreaded applications, and Shanghai is meant to improve singlethreaded. They were right then, I see no reason why their "20% IPC" promise should be invalidated now.
Edited by darkcloud89 - 6/17/08 at 9:07pm
post #29 of 55
Quote:
Originally Posted by darkcloud89 View Post
Really? Nehalem is by all means meant to be a server processor. The improvements made over Conroe/Penryn were meant to improve multithreaded performance and the IMC was to allow it to scale better beyond 1 and 2 socket systems. The Opteron by far outperforms Xeons in servers with more sockets. Intel is not the leader here. Barcelona was created to compete in this space, and Nehalem is the answer to Barcelona. Their respective architectures on desktop machines are irrelevant, it means nothing in comparison to server loads.



First of all, I already stated that AMD's manufacturing capacity is inferior. That is why it would have been stupid of them to counter Core2 on the desktop because Intel would still be manufacturing more and could sell them cheaper. This has nothing to do with the relative performance or significance of the underlying architecture.

I think a process revision will help AMD a lot more than you might realize. Designing a monolithic quad core on a 65nm process may have not been the best idea, it really is too big. Clock speed is arguably Phenom's biggest problem right now due to thermal management and power requirements. A die shrink to 45nm will help this and allow it to clock higher. HKMG will further alleviate the problem allowing it to clock even higher. This should come right around the same time as the 20% IPC improvement with Shanghai. I'm pretty confident that a 20% per-clock improvement as well as dramatic clock speed increase would put Shanghai higher in single-threaded performance than Penryn. Per-clock Phenom almost matches Penryn already, any IPC improvement should take it farther.



The improvements in Nehalem were mostly all done with regard to scaling and multithreading. Yes the benchmarks did show 40% improvement compared to Core2, but in what context was this? That's right, it was multithreading. What happened when Nehalem was compared to Penryn using only a single thread?

I disagree with this.

First of all, Nehalem is designed for application improvements from servers to notebooks. The QPI and monolithic die is as much for multi-socket bandwidth as it is for their integrated GPU project.

You did not answer my questions about where you assessment of 20-30% improvement in clock-for-clock performance and speeds starting at 3Ghz came from.

You raise a valid point that the performance improvements, clock for clock in a single L2 thread is minimal. However you seem to suggest that multithreading performance is only desired and beneficial to the server market?

What I find interesting is your assessment of Barcelona's supurior performance in 4 socket server market, because it isn't true:








Now please entertain us with a source for the 20-30% IPC inprovement and the 3.0Ghz+ launch clocks.
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Core i7 2500k ASRock P67 Extreme4 Gen 3 AMD 7970 16GB DDR3 
Hard DriveOptical DriveOSMonitor
Intel 520 256GB SATA DVD Burner Windows 7 64 bit Deal U2410 
KeyboardPowerMouse
Adesso Mechanical Silverstone OP650 Logitech G700 
  hide details  
Reply
post #30 of 55
The last time i checked Amd clearly said that they're not competing with Intel for performance... Penryn should be enough to take on Shanghai Nehalem is just too far for Amd to catch up even with Bulldozer i doubt it would even come close.
My System
(14 items)
 
  
CPUMotherboardGraphicsRAM
Intel Q9550 2.83GHz Asus P5Q Deluxe MSi N580 Twin Frozr III Lightning Extreme F2-8800CL5D-4GBPI 
CoolingOSMonitorPower
Thermalright Venomous X Window 7 x64 1080p TV monitor Corsair 1000watts 
Case
Naked 
  hide details  
Reply
My System
(14 items)
 
  
CPUMotherboardGraphicsRAM
Intel Q9550 2.83GHz Asus P5Q Deluxe MSi N580 Twin Frozr III Lightning Extreme F2-8800CL5D-4GBPI 
CoolingOSMonitorPower
Thermalright Venomous X Window 7 x64 1080p TV monitor Corsair 1000watts 
Case
Naked 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [TechReport] Intel gives peek into Nehalem bag of tricks