trthing
09-17-2003, 01:45 AM
http://asia.reuters.com/newsArticle.jhtml?type=technologyNews&storyID=3456448
Is this good for us CGers? Calling all the hardware savants!
Is this good for us CGers? Calling all the hardware savants!
View Full Version : Intel to Release Processor for PC Gaming
|
|
Valkyrien
09-17-2003, 02:05 AM
what are current standards for the cache? 512K? seems like that's the only thing that's really changed about these new processors...think about it, a processor with 4 times the standard cache size...but then again I'll have to wait for someone to explain exactly what the benefits of a larger processor cache would be;)
Goon
09-17-2003, 02:16 AM
I'm assuming 2 megs more will be 2.5 megs.
Isn't the cache the on processor equivelent to ram? if so this would mean even faster speeds, simply because the processor does not have to go to ram as often. thats a rather simple supposition, but beyond that i can't guess what impact it might have.
Will there be any specialized components actually for gaming in the chip?
Isn't the cache the on processor equivelent to ram? if so this would mean even faster speeds, simply because the processor does not have to go to ram as often. thats a rather simple supposition, but beyond that i can't guess what impact it might have.
Will there be any specialized components actually for gaming in the chip?
DeathBrain
09-17-2003, 02:52 AM
Great article, thanks for info :)
trthing
09-17-2003, 03:02 AM
You are welcome! I wish I knew enough to foresee the consequences of this for our job:shrug:
DeathBrain
09-17-2003, 03:35 AM
For our job ? imho...for realtime modeling & playback maybe, i saw big amount of cache ram there :)
3.2 gigahertz and come with an additional 2 megabytes of cache, sounds really cool for realtime playback :cool:
3.2 gigahertz and come with an additional 2 megabytes of cache, sounds really cool for realtime playback :cool:
trthing
09-17-2003, 03:49 AM
Hmmm I see... I thought it could have a special impact on render times...
By the way, I saw on another place that this launch could indicate some problems with Intel's 64-bit beast, Prescott...
By the way, I saw on another place that this launch could indicate some problems with Intel's 64-bit beast, Prescott...
DeathBrain
09-17-2003, 04:27 AM
Originally posted by trthing
Hmmm I see... I thought it could have a special impact on render times...
Of course there is an impact for rendering times, but not much :)
Large amount of processor cache helps us with preload/load large models or high resolution textures, woshh..big impact there, just like Xeon based processor :drool:
Hmmm I see... I thought it could have a special impact on render times...
Of course there is an impact for rendering times, but not much :)
Large amount of processor cache helps us with preload/load large models or high resolution textures, woshh..big impact there, just like Xeon based processor :drool:
Array
09-17-2003, 04:59 AM
how is this any differant from a xeon with a lot of cache?
DeathBrain
09-17-2003, 09:07 AM
I am still looking for those new processor white papers, i still dont know those 2 megabytes of cache, is it Execution Trace Cache (Level 1) or Advanced Transfer Cache (Level 2) or both.
Xeon uses 512KB to 1MB as their Advanced Transfer Cache
Still confused with this new processor, why the target is game market ? :hmm:
Xeon uses 512KB to 1MB as their Advanced Transfer Cache
Still confused with this new processor, why the target is game market ? :hmm:
alianor
09-17-2003, 11:20 AM
There is an article about this new processor here : http://www.anandtech.com/IT/showdoc.html?i=1870
In fact the 2MB more cache is L3 cache. I am not an expert, but I think I have read once that larger cache don't have a big inpact on rendering time.
Concering Real Time 3D it's another story. I think it's very important by reducting latencies when accessing to main memory.
Nevertheless more cache can't hurt. Only your wallet :p
For info I think that Athlon XP/MP has 128kb L1 & 512 KB L2, Opteron has 128kb L1 & 1Mb L2, P4 has ? L1 & 512 L2, Xeon DP has ? L1 & 512 kb L2 & 2MB L3.
So this new processor will be some kind of "gaming" Xeon DP, without 2 cpu possibilities of course.
In fact the 2MB more cache is L3 cache. I am not an expert, but I think I have read once that larger cache don't have a big inpact on rendering time.
Concering Real Time 3D it's another story. I think it's very important by reducting latencies when accessing to main memory.
Nevertheless more cache can't hurt. Only your wallet :p
For info I think that Athlon XP/MP has 128kb L1 & 512 KB L2, Opteron has 128kb L1 & 1Mb L2, P4 has ? L1 & 512 L2, Xeon DP has ? L1 & 512 kb L2 & 2MB L3.
So this new processor will be some kind of "gaming" Xeon DP, without 2 cpu possibilities of course.
Thalaxis
09-17-2003, 04:07 PM
Originally posted by Goon
I'm assuming 2 megs more will be 2.5 megs.
No, it seems that the cache is inclusive; that means that the
contents of L2 are mirrored in L3 (it simplifies coherency logic to
do things this way).
Isn't the cache the on processor equivelent to ram? if so this would mean even faster speeds, simply because the processor does not have to go to ram as often. thats a rather simple supposition, but beyond that i can't guess what impact it might have.
The cache is far, far faster than RAM. It takes a P4 over 150 clock
cycles to access main memory, while the L3 cache is probably
more like 20-30. Contrast this with Opteron which has a memory
access latency of just under 80 clock cycles, and you can start to
see why the Opteron platforms have been showing better
memory performance than P4's, and are only going to get better.
Will there be any specialized components actually for gaming in the chip?
I don't think so, that's coming in Prescott, due to start shipping in
December.
I'm assuming 2 megs more will be 2.5 megs.
No, it seems that the cache is inclusive; that means that the
contents of L2 are mirrored in L3 (it simplifies coherency logic to
do things this way).
Isn't the cache the on processor equivelent to ram? if so this would mean even faster speeds, simply because the processor does not have to go to ram as often. thats a rather simple supposition, but beyond that i can't guess what impact it might have.
The cache is far, far faster than RAM. It takes a P4 over 150 clock
cycles to access main memory, while the L3 cache is probably
more like 20-30. Contrast this with Opteron which has a memory
access latency of just under 80 clock cycles, and you can start to
see why the Opteron platforms have been showing better
memory performance than P4's, and are only going to get better.
Will there be any specialized components actually for gaming in the chip?
I don't think so, that's coming in Prescott, due to start shipping in
December.
Thalaxis
09-17-2003, 04:10 PM
Originally posted by trthing
Hmmm I see... I thought it could have a special impact on render times...
By the way, I saw on another place that this launch could indicate some problems with Intel's 64-bit beast, Prescott...
Prescott is a P4 on steroids, not a 64-bit chip. It's still pretty much
on schedule; what this shows is instead that Intel is looking to
match the Athlon64 FX at launch, which will have a 1 MB L2 and a
dual-channel memory controller. I think it will be launching at
2.0 and 2.2 GHz, or maybe the FX version will only be 2.2 and the
Athlon64 is going to be available at 2.0 and 2.2. Erm... well, you
get the idea.
Hmmm I see... I thought it could have a special impact on render times...
By the way, I saw on another place that this launch could indicate some problems with Intel's 64-bit beast, Prescott...
Prescott is a P4 on steroids, not a 64-bit chip. It's still pretty much
on schedule; what this shows is instead that Intel is looking to
match the Athlon64 FX at launch, which will have a 1 MB L2 and a
dual-channel memory controller. I think it will be launching at
2.0 and 2.2 GHz, or maybe the FX version will only be 2.2 and the
Athlon64 is going to be available at 2.0 and 2.2. Erm... well, you
get the idea.
Thalaxis
09-17-2003, 04:11 PM
Originally posted by Array
how is this any differant from a xeon with a lot of cache?
how is this any differant from a xeon with a lot of cache?
Thalaxis
09-17-2003, 04:17 PM
Originally posted by DeathBrain
I am still looking for those new processor white papers, i still dont know those 2 megabytes of cache, is it Execution Trace Cache (Level 1) or Advanced Transfer Cache (Level 2) or both.
Xeon uses 512KB to 1MB as their Advanced Transfer Cache
Still confused with this new processor, why the target is game market ? :hmm:
Advanced transfer cache: just fast cache with a silly marketing
name. Other than being faster than anyone else's cache, it's not
anything different. (Well, Madison's cache might be faster, but it's
still an Intel dealie, so there you go.)
Execution Trace Cache: equivalent to everyone else's L1 instruction cache, but unique to the P4's oddball method; it
decouples the instruction decoder from the critical timing path,
which makes it easier to ramp clock speeds, and allows the
scheduler to deal with finding instruction-level parallelism with
decoded micro-ops, rather than the method AMD used which is 3
hugely complicated instruction decoders that run in parallel.
Both are impressive technical achievements, don't get me wrong;
they are just highly reflective of the different design philosophies
at work in the two processors.
Intel has a separate data cache, which is 8 kB in the Northwood,
16 in Prescott, and insanely fast. They traded off speed for size
there, and made up for it with prefetch and an insanely fast L2.
I am still looking for those new processor white papers, i still dont know those 2 megabytes of cache, is it Execution Trace Cache (Level 1) or Advanced Transfer Cache (Level 2) or both.
Xeon uses 512KB to 1MB as their Advanced Transfer Cache
Still confused with this new processor, why the target is game market ? :hmm:
Advanced transfer cache: just fast cache with a silly marketing
name. Other than being faster than anyone else's cache, it's not
anything different. (Well, Madison's cache might be faster, but it's
still an Intel dealie, so there you go.)
Execution Trace Cache: equivalent to everyone else's L1 instruction cache, but unique to the P4's oddball method; it
decouples the instruction decoder from the critical timing path,
which makes it easier to ramp clock speeds, and allows the
scheduler to deal with finding instruction-level parallelism with
decoded micro-ops, rather than the method AMD used which is 3
hugely complicated instruction decoders that run in parallel.
Both are impressive technical achievements, don't get me wrong;
they are just highly reflective of the different design philosophies
at work in the two processors.
Intel has a separate data cache, which is 8 kB in the Northwood,
16 in Prescott, and insanely fast. They traded off speed for size
there, and made up for it with prefetch and an insanely fast L2.
CGTalk Moderation
01-16-2006, 03:00 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.
vBulletin v3.0.5, Copyright ©2000-2012, Jelsoft Enterprises Ltd.