corvetteguy
Distinguished
- Jan 15, 2006
- 1,545
- 0
- 19,780
- Feb 1, 2006
- 2,034
- 1
- 19,780
@Piddy: This is in my "Bookmarks" in my Opera Browser.
@Chime: I'm going to try and explain what I think you you're asking:
The reason companies use a longer Pipeline (I.E. Intel) is because the longer the pipeline, the faster you can crank the Internal Clock Speed. Intel's marketing department was chasing after the Athlon XP and on a Clock-For-Clock basis, the Pentium 4 was suffering. The Athlon XP has a 10-Stage Pipeline, whereas the Pentium 4 had between 15 and 20 (nowawdays, a Pentium 4 has up to 31-Stages). Intel decided they needed to aggressively increase the clock speed because they couldn't match the Athlon XP, so they kept increasing the size of the pipeline and made promises like "10GHz Pentium 4", which is impossible under circumstances due to trimense heat.
On the Compiler part, I think you're referring to when Hazards happen. A Hazzard is when multiple instructions are relying on each other for data and when 1 finishes first, it can get wrong data if the instruction it relies on is not done being executed. Other Hazards such as when parts of the CPU need 2 or more Instructions, and they both required the ALU, that can cause a Structural Hazard. There are ways that are implemented to reduce Hazards, such as Flushing The Pipeline and using NOP's (Non OPeration). When the CPU thinks there might be a problem or a Hazard, it inserts these NOP instructions into the pipeline. An NOP does absolutely nothing at all other than take up space for a duration of Clock Cycles.
I hope that answered your question, if it didn't, repost and I will try to accomodate you.
~~Mad Mod Mike, pimpin' the world 1 rig at a time
ches111
Distinguished
- Feb 4, 2006
- 1,958
- 0
- 19,780
Hey Mike,
Good thread with bias
.
Just a FYI for those still pondering the 32bit vs 64bit questions.
Mike was correct in stating registers. There are roughly three main characteristics that need to be taken into account when deciphering 32 vs 64 bit.
1) Registers -- The registers mike was talking about need typically three steps to complete a single transaction. There is a Load, Execute and Clear transaction for each register with multiple registers in operation. These registers are typically the size of the XXbit processors specs. I say typically because some vendors tried to release psuedo 64bit architectures (two linked 32bit registers somewhat acting as a single 64). So higher clock cycles would allow for a thoeretically quciker Load, Execute and Clear. This would assume more work in the same time. This of course does not fully include the pipeline information which could be larger or smaller based upon design. These allow for more or less operations to occur at each STEP (greater number of steps the less work typically at each step and vice versa).
2) Memory Footprint -- The reason 64bit Procs have been underused is that they have to have mods to their software to take advantage of the procs characteristics. Going from 32 to 64 requires you to be smarter with usage of memory. If you can have for instance access to a 32 bit INT or an even bigger LONG of 64bit. When memory is reserved in code you now reserve a larger address. If you only use say 4b of the address you are effectively wasting the rest of the address.
3) Word Alignment -- Some people will know and understand that word aligning on systems that are 64 bit is crucial. Bit shifting in a 64 bit architecture is DEATH. You could actually slow the system to slower than its same 32bit counterpart. Since we now have to word align (typically 8bit alignment in most 64bit arch) we get REAL sloppy with memory handling/usage (sometimes padding things to word align). Programs get larger and if not done well also could be much slower. This optimization has been mostly reserved for servers and enterprise software since memory requirements could get large.
Some additional info Here.
Hope this only contributes to this thread.
Good thread with bias
.Just a FYI for those still pondering the 32bit vs 64bit questions.
Mike was correct in stating registers. There are roughly three main characteristics that need to be taken into account when deciphering 32 vs 64 bit.
1) Registers -- The registers mike was talking about need typically three steps to complete a single transaction. There is a Load, Execute and Clear transaction for each register with multiple registers in operation. These registers are typically the size of the XXbit processors specs. I say typically because some vendors tried to release psuedo 64bit architectures (two linked 32bit registers somewhat acting as a single 64). So higher clock cycles would allow for a thoeretically quciker Load, Execute and Clear. This would assume more work in the same time. This of course does not fully include the pipeline information which could be larger or smaller based upon design. These allow for more or less operations to occur at each STEP (greater number of steps the less work typically at each step and vice versa).
2) Memory Footprint -- The reason 64bit Procs have been underused is that they have to have mods to their software to take advantage of the procs characteristics. Going from 32 to 64 requires you to be smarter with usage of memory. If you can have for instance access to a 32 bit INT or an even bigger LONG of 64bit. When memory is reserved in code you now reserve a larger address. If you only use say 4b of the address you are effectively wasting the rest of the address.
3) Word Alignment -- Some people will know and understand that word aligning on systems that are 64 bit is crucial. Bit shifting in a 64 bit architecture is DEATH. You could actually slow the system to slower than its same 32bit counterpart. Since we now have to word align (typically 8bit alignment in most 64bit arch) we get REAL sloppy with memory handling/usage (sometimes padding things to word align). Programs get larger and if not done well also could be much slower. This optimization has been mostly reserved for servers and enterprise software since memory requirements could get large.
Some additional info Here.
Hope this only contributes to this thread.
ches111
Distinguished
- Feb 4, 2006
- 1,958
- 0
- 19,780
On Hazzards,
Developers will typically LOCK via mutex data that is shared between multiple threads. This will avoid many of those collisions at runtime. In any multi-threaded application you stand the chane of data synchronization and corruption. These locks (when implemented correctly) help to alleviate these issues. This is done to help get over the hardware issues.
Most procs break themselves up into four distinct sectors (Process, Kernel, System and User). This allows for context switching. While in one quadrant of the proc you need to block and interupt to allow the next quadrant to process. The multi-thread HT in the newer Intel P 955 will allow for four of these quadrant distinct areas (four threads per cycle). This is important becasue you can context switch four independent threads per cycle. This will become a possible advantage if for instance software is written optimizing its use. That is why Intel is having the devolpers conference and releasing info about its new processors at the same time. I am sure a tool kit will be released to them. This is why it is easy to BEWARE benchmarks.
A good book for reference is "UNIX INTERNALS the new Frontiers".
Great book showing the interaction between the proc and software written for it. Of course this is UNIX bias.
Developers will typically LOCK via mutex data that is shared between multiple threads. This will avoid many of those collisions at runtime. In any multi-threaded application you stand the chane of data synchronization and corruption. These locks (when implemented correctly) help to alleviate these issues. This is done to help get over the hardware issues.
Most procs break themselves up into four distinct sectors (Process, Kernel, System and User). This allows for context switching. While in one quadrant of the proc you need to block and interupt to allow the next quadrant to process. The multi-thread HT in the newer Intel P 955 will allow for four of these quadrant distinct areas (four threads per cycle). This is important becasue you can context switch four independent threads per cycle. This will become a possible advantage if for instance software is written optimizing its use. That is why Intel is having the devolpers conference and releasing info about its new processors at the same time. I am sure a tool kit will be released to them. This is why it is easy to BEWARE benchmarks.
A good book for reference is "UNIX INTERNALS the new Frontiers".
Great book showing the interaction between the proc and software written for it. Of course this is UNIX bias.
- Feb 1, 2006
- 2,034
- 1
- 19,780
I don't specialize in programming (I do DirectX Windows Programming sometimes) but rather strictly General Hardware, as you can see from my 1st post. But thanks for adding, even though this was geared towards people with little understanding
. I actually might move away from computers and try to become a Police Detective, I like helping people and teaching rather than working in a job such as Programming, so maybe even become a Teacher, heh.~~Mad Mod Mike, pimpin' the world 1 rig at a time
ches111
Distinguished
- Feb 4, 2006
- 1,958
- 0
- 19,780
Actually Mike,
Too often people want to separate Hardware and Software. One is worthless without the other.
@ the time the Dec Alpha chips were released they had 200+mhz chips that were 64 bit. People had no idea what to do with them. They wrote poor applications that would cripple the hardware and then state that the procs were horrible in performance. A simple alignment of words in the software would give a generous performance increase over its predecessors.
They are so so linked especially Operating Systems!
Too often people want to separate Hardware and Software. One is worthless without the other.
@ the time the Dec Alpha chips were released they had 200+mhz chips that were 64 bit. People had no idea what to do with them. They wrote poor applications that would cripple the hardware and then state that the procs were horrible in performance. A simple alignment of words in the software would give a generous performance increase over its predecessors.
They are so so linked especially Operating Systems!
- Feb 1, 2006
- 2,034
- 1
- 19,780
True, but I'm not making the hardware lol. I just learn about it and inform others who don't know, if they care enough to need to know how to program for it, so be it.
~~Mad Mod Mike, pimpin' the world 1 rig at a time
- Feb 1, 2006
- 2,034
- 1
- 19,780
ok mike sorry i didnt know geez and man bribe fredi to make this a sticky
Jack Barnes said he conversed with all the other Moderators and they concluded that they wouldn't make it a Sticky, and I won't bribe or beg anybody (not sure how I can bribe someone over the internet).
~~Mad Mod Mike, pimpin' the world 1 rig at a time
ches111
Distinguished
- Feb 4, 2006
- 1,958
- 0
- 19,780
What Mike,
Your not striving for STICKYNESS? What about the strawberries?
Man can you imagine going your whole life without having ONE sticky thread?
Man the shame that would bring on your family!!
Mike gets sticky:
I'd like to give thanks to GOD and my country, and you know I love ya MOM. I still have hope for world piece!! (OSCARs combined with Miss America)
Your not striving for STICKYNESS? What about the strawberries?
Man can you imagine going your whole life without having ONE sticky thread?
Man the shame that would bring on your family!!
Mike gets sticky:
I'd like to give thanks to GOD and my country, and you know I love ya MOM. I still have hope for world piece!! (OSCARs combined with Miss America)
- Feb 1, 2006
- 2,034
- 1
- 19,780
What Mike,
Your not striving for STICKYNESS? What about the strawberries?
Man can you imagine going your whole life without having ONE sticky thread?
Man the shame that would bring on your family!!
Mike gets sticky:
I'd like to give thanks to GOD and my country, and you know I love ya MOM. I still have hope for world piece!! (OSCARs combined with Miss America)
HaHa you can't sell PEACE
.I don't want sticky, this thread is too good for a sticky, it's 1337 :twisted:.
BTW, you got any strawberries yet? I'm really hungry
.~~Mad Mod Mike, pimpin' the world 1 rig at a time
- Feb 1, 2006
- 2,034
- 1
- 19,780
ak47is1337
Distinguished
- Jan 30, 2006
- 1,830
- 0
- 19,780
wheres the rundown on nvidia/ati?........eh?Id like to know why my x1600 pro runs so hot.....I had to mod the stock heatsink and redo my case a bit to get my gpu to idle at 45c.(Cooling on the cheap)
windows help reduce case temps,just a little.or it was just an excuse to whack my case.I think plastic releases heat faster than metal.
i put an extra fan on my gpu heat sink.stock temps were way high the heat sink cover was over 100 f . now its at 27c in the heatsink fins.
why am i reading stock temps are so high ? the nvidias are high too.
speed is good ,but heat wears down your stuff,what the lifespan on a stock 7800gtx or an x 1900xt with that kind of heat?
the 7800gtx has more heat than my card at idle and load.I should mention my card is oc'd with atitool at 537 and 432.and my case is pushing 150 cfm(2x75cfm) in from the fans on the side out from the top(2x20cfm )and back(1x20+1x60cfm),and psu(2x20cfm=20cfm ttl).=pos.case pressure of 10+/-? cfm
anyway those high temps are bad....right(stock temps)
Who cares? It should be warranted. Also, GPU's can withstand a lot of heat, although it certainly isn't good for overclocking and I can't see it doing good for PCB lifetime, it can't do too terribly much either.
- Feb 1, 2006
- 2,034
- 1
- 19,780
wheres the rundown on nvidia/ati?........eh?Id like to know why my x1600 pro runs so hot.....I had to mod the stock heatsink and redo my case a bit to get my gpu to idle at 45c.(Cooling on the cheap)
windows help reduce case temps,just a little.or it was just an excuse to whack my case.I think plastic releases heat faster than metal.
i put an extra fan on my gpu heat sink.stock temps were way high the heat sink cover was over 100 f . now its at 27c in the heatsink fins.
why am i reading stock temps are so high ? the nvidias are high too.
speed is good ,but heat wears down your stuff,what the lifespan on a stock 7800gtx or an x 1900xt with that kind of heat?
the 7800gtx has more heat than my card at idle and load.I should mention my card is oc'd with atitool at 537 and 432.and my case is pushing 150 cfm(2x75cfm) in from the fans on the side out from the top(2x20cfm )and back(1x20+1x60cfm),and psu(2x20cfm=20cfm ttl).=pos.case pressure of 10+/-? cfm
anyway those high temps are bad....right(stock temps)
Well, I'm sure there is some nice article in Graphics like mine here for CPU's
My 6800GS's idle about 39c - 45c, which is normal. Load temps go to 55c to 72c, again, which is normal. Your GPU's are going to be alot hotter than your CPU.
~~Mad Mod Mike, pimpin' the world 1 rig at a time
- Feb 1, 2006
- 2,034
- 1
- 19,780
I for one would like to secure the life of an investment ,sonny. 8)
AFAIK, temps recorded to Windows are of the entire card, meaning RAM and GPU. If you have Memory running at over 1GHz to 1.6GHz and a GPU of 600MHz right next to each other, things are going to get quite toasty
.~~Mad Mod Mike, pimpin' the world 1 rig at a time
- Feb 1, 2006
- 2,034
- 1
- 19,780
You're asking about CPU temps and why yours are hot? Do you have a presshot?
~~Mad Mod Mike, pimpin' the world 1 rig at a time
- Feb 1, 2006
- 2,034
- 1
- 19,780
LOL!
Dude, you aint gonna get your GPU as cool as your CPU, at least under load. But, if you were a Mad Modder Like me, you could record these temps like I did
:CPU:
Stock Heatsink: 33c Load, 27c Idle
GPU:
Stock Heatsink: 35c Idle 45c Load
~~Mad Mod Mike, pimpin' the world 1 rig at a time
- Feb 1, 2006
- 2,034
- 1
- 19,780
verndewd said:
LOL!
Dude, you aint gonna get your GPU as cool as your CPU, at least under load. But, if you were a Mad Modder Like me, you could record these temps like I did
CPU:
Stock Heatsink: 33c Load, 27c Idle
GPU:
Stock Heatsink: 35c Idle 45c Load
dont break your arm patting yourself mike,i am a moddin freakazoid.when my temps go down and match yours ,ill send ya some gpu roasted crow to eat.
LOL, iight man, but don't forget, I'm the self proclaimed best modder on Air, my best was fitting a Dual AMD Opteron 64 Workstation w/ 12 120mm Fans and 30 80mm Fans. It dropped temps by over 15c in some scenarios.
~~Mad Mod Mike, pimpin' the world 1 rig at a time
- Feb 1, 2006
- 2,034
- 1
- 19,780
Well, when you drop 15 from 40c normal with 8 Hard Drives, over 40 cables, 2 CPU's and a host of additional devices, that's quite an achievement in my book.
~~Mad Mod Mike, pimpin' the world 1 rig at a time
- Feb 1, 2006
- 2,034
- 1
- 19,780
LOL, nah. I'll go into MSPAINT right now and draw a diagram of it and post it. Gimme about 5 minutes, it's worth the wait
.~~Mad Mod Mike, pimpin' the world 1 rig at a time
ak47is1337
Distinguished
- Jan 30, 2006
- 1,830
- 0
- 19,780
I for one would like to secure the life of an investment ,sonny. 8)
I hate to break it to you "sonny" but if you really want to secure your investment you might as well just have the company replace it as that's guaranteed whereas putting an aftermarket cooler might invalidate the warranty and you could still get screwed.
- Feb 1, 2006
- 2,034
- 1
- 19,780
http://img99.imageshack.us/my.php?image=workstation1kq.jpg
It isn't the best but that's the setup I have, excluding the 10 used for exhaust.
~~Mad Mod Mike, pimpin' the world 1 rig at a time
It isn't the best
but that's the setup I have, excluding the 10 used for exhaust.~~Mad Mod Mike, pimpin' the world 1 rig at a time
- Feb 1, 2006
- 2,034
- 1
- 19,780
I should mention the case is one that is basically 2 stuck together, it is designed for workstations/servers. I think it'd be interesting for somebody with the right funds to test the best OC on Air.
~~Mad Mod Mike, pimpin' the world 1 rig at a time
~~Mad Mod Mike, pimpin' the world 1 rig at a time
Facebook
Twitter
Instagram