[navin]

Originally Posted by StarScream

After my ten year old P4 stopped booting, decided to get a new machine from Dell. Placed an order for a Dell XPS 7100 earlier this week:
AMD 1055T (6 cores, 2.8GhZ)
6GB ddr3 RAM
Two hard drives (750 + 500GB)
ATI 5670 graphics card
Windows 7 64-bit professional
23" Dell monitor

6 cores? wow! What did it cost.

Another Newbie question (dont laugh - I am old and much of what I know is not relavant).

When you ahve a multiple core processor does not the software (applications, shells, etc..) have to be writtent o manage these multiple cores and processes or does Intel / AMD embedd their processors with a mutli-core manager? For example if I were to run a FEA software like IDEAS http://www.ideas-eng.com/finite_element.html would it run faster on a 4 /6 core machine even if it was written for a 2 core processor?

[anku94]

Originally Posted by navin

6 cores? wow! What did it cost.

The processor itself is quite cheap for the number of cores, $285 as per AnandTech. In comparison, Intel's 6-core Core i7 980x costs $1000. It's obviously better than the AMD though, but you can't beat AMD on performance/price ratio.

Originally Posted by navin

When you ahve a multiple core processor does not the software (applications, shells, etc..) have to be writtent o manage these multiple cores and processes or does Intel / AMD embedd their processors with a mutli-core manager? For example if I were to run a FEA software like IDEAS IDEAS Ltd - Finite Element Analysis would it run faster on a 4 /6 core machine even if it was written for a 2 core processor?

It's the software that has to be optimized to be used on multiple cores. If you're running, say, 10 different programs, the processor will distribute them evenly, but if you're running one big program, it's the program that has to be optimized. So, I think, I'm not sure though, the quad-core might be able to provide slight performance gains because of its architecture, but unless a software is optimized, the gains would be quite less.

[StarScream]

Originally Posted by navin

6 cores? wow! What did it cost.

less than Intel's 4-core i5-750 by a fair margin. That's why I went for it. Total system cost minus monitor was 70k

Another Newbie question (dont laugh - I am old and much of what I know is not relavant).

When you ahve a multiple core processor does not the software (applications, shells, etc..) have to be writtent o manage these multiple cores and processes or does Intel / AMD embedd their processors with a mutli-core manager? For example if I were to run a FEA software like IDEAS IDEAS Ltd - Finite Element Analysis would it run faster on a 4 /6 core machine even if it was written for a 2 core processor?

Originally Posted by anku94

The processor itself is quite cheap for the number of cores, $285 as per AnandTech. In comparison, Intel's 6-core Core i7 980x costs $1000. It's obviously better than the AMD though, but you can't beat AMD on performance/price ratio.



It's the software that has to be optimized to be used on multiple cores. If you're running, say, 10 different programs, the processor will distribute them evenly, but if you're running one big program, it's the program that has to be optimized. So, I think, I'm not sure though, the quad-core might be able to provide slight performance gains because of its architecture, but unless a software is optimized, the gains would be quite less.

As anku94 says, the applications have to be optimized for multicore processors, otherwise there is little benefit. The four core Intel performs better in games because there are hardly any games that are multi-threaded. I don't want this for gaming though. More business use with many apps at the same time.
AMD just can't be beat for value and price/performance. Im going to keep this machine for a while so six-cores was the better way to go IMHO.

[IronH4WK]

Originally Posted by StarScream

The four core Intel performs better in games because there are hardly any games that are multi-threaded.

yep, that's true. i've got a AMD Phenom X4 (quad-core) and when running games, only one of the cores is utilized (games like the Battlefield series, Crysis etc.) whereas running something like the Adobe Premiere Pro will put a load all the four cores.

multi-core optimized software perform better than their counterparts. also, i think (not sure) the architecture (32/64 bit) also matters. i've noticed that a 64-bit Photoshop performs way faster than the 32-bit version.

[StarScream]

Forgot to mention, the 70K also includes 3-year Dell warranty. That alone adds 11k to the price.

[navin]

Originally Posted by StarScream

More business use with many apps at the same time.

If you are running multiple applications does he OS then handle the tasking? With 6 cores someone must be telling the cores what to do and when na? Do the 6 cores run different clocks? can they? if they do/could it'd would be a killer na?

What is the die size? What is the internal cache?

Originally Posted by StarScream

Forgot to mention, the 70K also includes 3-year Dell warranty. That alone adds 11k to the price.

Nice 70k for 6 cores is nice esp with a 3 year warranty. You can't get a i5-750/8GB based macine from Intel for that money.

[anku94]

Originally Posted by navin

If you are running multiple applications does he OS then handle the tasking? With 6 cores someone must be telling the cores what to do and when na? Do the 6 cores run different clocks? can they? if they do/could it'd would be a killer na?

What is the die size? What is the internal cache?

• 768KB L1 Cache (Instruction + Data): 128KB x6 (64KB + 64KB for each core)
• 3MB L2 Cache: 512KB x6 (quad-core)
• 6MB L3 Cache: 6MB Shared L3

Die size : 346 mm^2



AFAIK, all cores run at the same clock but Intel has introduced a new tech called turbo-boost which temporarily overclocks one or two cores if you're running a high single-threaded application.

And the allocation of cores to programs is handled by the processor itself. Which is why older OSes can be run, I think. Although modern OSes may optimize the process to some extent, but largely it's the processor's job.

[Prowler]

Finally yesterday I made the purchase.

Low end Computer:

Intel Atom CPU with Mother board
2 GB DDR2 RAM
500 GB Seagate HD
15.6 LCD Monitor

Total cost: about Rs 15000/-

High End Computer:

Intel i5-760
Intel DH55TC Motherboard
Corsair 2GB x 2 DDR3 RAM
SeaGate 500 GB HD
Logitech Wireless KBD/Mouse
Primesource Cabinet + SMPS
Zotec GF 8400 GS 512 MB Graphics card

Cost: Rs.35000/

Windows Home Basic:Rs4300/-

I plan to move the existing Windows XP into the new highend computer. As this MB doesn't support IDE, I would have to clone the IDE drive into SATA.
Clonezilla should do the trick.

Thanks Aggoswami and Anku94 for all your valuable inputs. I opted for intel for one reason. My present computer never gave me any problem. We have 4 intel and 10 AMD in our office. Out of this, 4 AMD had failed whereas none of the Intel failed. I didn't want to take a chance on my workhorse.

[NetfreakBombay]

Originally Posted by navin

When you ahve a multiple core processor does not the software (applications, shells, etc..) have to be writtent o manage these multiple cores and processes or does Intel / AMD embedd their processors with a mutli-core manager? For example if I were to run a FEA software like IDEAS IDEAS Ltd - Finite Element Analysis would it run faster on a 4 /6 core machine even if it was written for a 2 core processor?

Yes applications have to be written to exploit multiple cores. And most heavy-duty number crunching softwares does support as many cores as are available.

IDEAS has excellent support for multiple core processors. And you can configure how many "threads of execution" it should run. To load a 6 core system to its limit, this number would be 6.

Originally Posted by navin

If you are running multiple applications does he OS then handle the tasking? With 6 cores someone must be telling the cores what to do and when na? Do the 6 cores run different clocks? can they? if they do/could it'd would be a killer na?

Yes OS handles the allocation and deallocation of CPU cores to various applications that are running at the same time.

In newer CPUs all 6 cores can run at different clocks, this is to save on electricity costs.

Originally Posted by navin

What is the die size? What is the internal cache?

Die size is actual surface area of CPU chip in mm square. Larger the die size, higher the cost. Since with a single waffer, fewer chips can be cut.

CPU -> memory communication is costly (CPU has to request something for memory and that core sleeps until Memory chips responds. Memory is MUCH slower than CPU)

To solve this problem, CPU has some memory of its own (cache) this memory runs at the same speed as CPU. So. frequently used data items can be stored in cache.

[StarScream]

Originally Posted by navin

If you are running multiple applications does he OS then handle the tasking? With 6 cores someone must be telling the cores what to do and when na? Do the 6 cores run different clocks? can they? if they do/could it'd would be a killer na?

What is the die size? What is the internal cache?



Nice 70k for 6 cores is nice esp with a 3 year warranty. You can't get a i5-750/8GB based macine from Intel for that money.

the 1055t is a 45nm chip. The latest version consumes 95w (an earlier version took 125w). Multiple apps won't automatically work on multiple cores but it does give you more headroom. As far as I know the six cores can't run at different speeds, except in turbo boost where 3 cores can go up to 3.3 GhZ. Internal cache is 9MB.

[navin]

Originally Posted by anku94

• 768KB L1 Cache (Instruction + Data): 128KB x6 (64KB + 64KB for each core)
• 3MB L2 Cache: 512KB x6 (quad-core)
• 6MB L3 Cache: 6MB Shared L3

Die size : 346 mm^2

Attachment 404394

AFAIK, all cores run at the same clock but Intel has introduced a new tech called turbo-boost which temporarily overclocks one or two cores if you're running a high single-threaded application.

Originally Posted by StarScream

the 1055t is a 45nm chip. The latest version consumes 95w (an earlier version took 125w). Multiple apps won't automatically work on multiple cores but it does give you more headroom. As far as I know the six cores can't run at different speeds, except in turbo boost where 3 cores can go up to 3.3 GhZ. Internal cache is 9MB.

45nm is the transistor gate (also called 'circuit' or more techically correct 'linewidth') dimension (based on the '45nm process') the die size is usually in mm in this case 364 mm (the physical surface area of the wafer) which might look rather large to some but even in the 80s we (IBM/AMD/Hitachi/Intel/Kyocera) already had 125mm dies in production and 150mm was around the corner. Transistor density however has gone up many fold in the past 25-30 years this density is what has raised the bar (the closer and smaller the transistors are not only can you fit many more on a chip but the circuits are faster - eventually we cant go faster than 186,000 miles per sec na?).


Chip Rant
I would rather CPUs be measured in FLOPS or some other standard that is more clear about their processing power than GHz and process technology. I mean what does the end user care if AMD or Intel claims that their processor has "16 cores" vs the compeition's "12 cores" and for heavens sake I have lost complete faith in the efficacy of "Interlagos", 'Lynnfield', "Opteron", "Bulldozer", etc..I come across in the Rags I once used to love to read (they were once good Mags).

I made a call to an old friend (I cant name him) Iused to work with and he told me that as far as real world operations go the 6 core i7-980 is only about 20% faster than the 2 core i5-650/655. BUMMER!

Reality Check Rant
Some 10 years ago I got me a PC with a Pentium chip with 180nm line width and a 224mm2 die size. I think it was 1.5Ghz or something. I had Windows 98SE and Office and wrote Excel sheets and Word docs. Today I have a PC that is 100x faster with a 45nm line width and 346mm die (or whatever I have not checked) but I still run Excel and Word and produce the same quality of documets I did in 1995 or 2000. I dont even do it any faster or slower. In between al these years I saw chips having 130nm//205mm2 then 90nm/112 mm2 then 65nm/190 mm2. My customers, colleagues, vendors did not notice any difference these processors (or later versions of Windows/Office) made to my output. I am sure that when I get a i9-1000 16 core 4.7Ghz processor with Windows 8 it still wont make any difference. I am the bottleneck.

Disclaimer
Yes I still maintain I know precious little about CPUs and stuff. Much of what I learnt in the 80s is not valid today. Thankfully electrcity still travels at 186,000 miles per sec (300,000 km per sec for the FPS-challendged)

[srishiva]

It all boils down to concurrency and if it was designed into the application. Not all applications have it designed in them or its not possible to achieve concurrency depending on the type of the job you are trying to accomplish.
Most of the time its energy efficiency vs performance that chip designers try to achieve. Is AMD multi-threaded too?

[anku94]

Originally Posted by navin

Chip Rant
I would rather CPUs be measured in FLOPS or some other standard that is more clear about their processing power than GHz and process technology. I mean what does the end user care if AMD or Intel claims that their processor has "16 cores" vs the compeition's "12 cores"......

But won't measurement in FLOPS highlight the best possible performance in an ideal situation ? I mean, in real world, you don't use all the cores for most of the time, and in such case, wouldn't the actual processing power be less than what's reflected in the FLOPS number ? Whereas cores give you a fair idea of what to expect.

Pardon me but it's my turn to be stupid now. I'm practically stupid when it comes to the actual technology of processors.

Originally Posted by navin

Reality Check Rant
Some 10 years ago I got me a PC with a Pentium chip with 180nm line width and a 224mm2 die size. I think it was 1.5Ghz or something. I had Windows 98SE and Office and wrote Excel sheets and Word docs. Today I have a PC that is 100x faster with a 45nm line width and 346mm die (or whatever I have not checked) but I still run Excel and Word and produce the same quality of documets I did in 1995 or 2000. I dont even do it any faster or slower. In between al these years I saw chips having 130nm//205mm2 then 90nm/112 mm2 then 65nm/190 mm2.

Try the same with Crysis.

Originally Posted by navin

....but even in the 80s we (IBM/AMD/Hitachi/Intel/Kyocera) already had 125mm dies in production and 150mm was around the corner.

Why does it take better technology to increase die size ? Don't you just have to cut wafers to a bigger size ?

[Thad E Ginathom]

Originally Posted by navin

I am the bottleneck.

Absolutely!

Same thing applies to Browser Wars, where how fast a page is drawn is measured in thousandths of a something. I don't care. If I don't have any perceived wait for the page, I am happy!

Most of the computing most of us do would be just fine on much lesser hardware. I wish I hadn't given away my last P4 machine: it would have made a perfectly fine Linux machine.

Speaking of which, the most huge difference in my computing experience in many years (going back to, it might have been, the first 486s, where we were amazed to find we couldn't actually read a DIR listing as it scrolled up the screen) has come from switching to Linux. My machine is now ready to use in one minute, not five. My browser is now ready to browse in a couple of seconds, not a minute or so. We may be the bottleneck, but Windows, in all of its various incarnations, has not helped.

[navin]

Originally Posted by anku94

But won't measurement in FLOPS highlight the best possible performance in an ideal situation ?

Try the same with Crysis.

Why does it take better technology to increase die size ? Don't you just have to cut wafers to a bigger size ?

Nope FLOPs is only the meaure of number of (Floating Point) instructions/sec but it is still a far better yardstick that Ghz, Cores, die size, etc.. the i7-980 mentioned in my earlier post for example is about 100 GigaFLops.

While FLOPS is in effect a measure of the computer's performance for vector maths (not scalar maths) it is vector maths that is used by most if not all DSP/graphics programs which is why a dedicated graphics machine like a XBOX/PS3 will have a higher FLOP rating that the i5/i7 series but in reality cant do as much as a i5/i7.

Ofcourse that geek from Seattle has an insatiable hunger for FLOPS and as Thad explained his products do a damn good job of making an i7 perform about as fast as a P4 running Linux.