It seems like Intel and AMD are going after having more cores than before and 8 cores is planned for next year AFAIK. It seems like clock speed has pretty much topped out at around 3GHZ since the Pentium 4 days and these two companies are creating better architectures and more cores in their chips, which can't be matched with pure overclocking as it could in past times. Would you say that Intel and AMD are entering a core race now?
Are AMD and Intel entering a core race?
- Thread starter Smartazz
- Start date
Ricemarine
Lifer
- Sep 10, 2004
- 10,507
- 0
- 0
The question is though... what's after increasing the number of cores? Reducing the nm? Then what's after that? :laugh:...
Although, it would be a while...
otherwise the engineers will have to find another jobA new way of developing processors will emerge.
Edit: Once CPU's and GPU's are on the same die. DUN DUN DUNNNNN
Edit: Once CPU's and GPU's are on the same die. DUN DUN DUNNNNN
AMD already said they won't go too far into the core race
They are looking into other areas with stuff like torrenza and fusion(they don't want to be the one creating the high core version of netburst,u know power hungry and slow)
They are looking into other areas with stuff like torrenza and fusion(they don't want to be the one creating the high core version of netburst,u know power hungry and slow)
I, for one, am not impressed with all the hoopla around anything with more than 4 cores or 3GHz.
what I want to see is cheaper prices, lower temps, and lots more cache on every level.
more low level cache always** brings more performance but its expensive.
** Yes, I know. I've studied 2 introductory courses in processor designs and suddenly I think I know everything. I am aware of all the "speed-up factor" calculations etc etc...But for home computing, this is still true.
what I want to see is cheaper prices, lower temps, and lots more cache on every level.
more low level cache always** brings more performance but its expensive.
** Yes, I know. I've studied 2 introductory courses in processor designs and suddenly I think I know everything. I am aware of all the "speed-up factor" calculations etc etc...But for home computing, this is still true.
For someone named JEDIYoda, you're hardly one to judge.
Use the force to grow a brain, and learn to be more polite to people.
i'll repeat myself so that even you could understand: more cache is always better for home computing = games, watching movies, surfing the web etc etc.
sadly, you too fail to see the point.
and what if silicon chips reach their frequency wall, what then? Intel and AMD close shop?
the company that will produce low cost, quad cores with good amounts of cache processors would win the home PC market. right now, it looks like Intel is on its way with the 266$ quad later this year.
think about how good would be a 32-core 10GHz processor with 1MB cache per core and i can already tell you the cache will be heavily overflowed to an extend it will cripple the processor.
the increase in frequency and number of cores has to be coupled with increase in cache.
and what if silicon chips reach their frequency wall, what then? Intel and AMD close shop?
the company that will produce low cost, quad cores with good amounts of cache processors would win the home PC market. right now, it looks like Intel is on its way with the 266$ quad later this year.
think about how good would be a 32-core 10GHz processor with 1MB cache per core and i can already tell you the cache will be heavily overflowed to an extend it will cripple the processor.
the increase in frequency and number of cores has to be coupled with increase in cache.
myocardia
Diamond Member
- Jun 21, 2003
- 9,291
- 30
- 91
I don't think so. My 2.0 Ghz dual-core would outperform any processor I've seen so far, including a 4 Ghz quad-core, using any application that I will ever use.
Intel is already developing a processor that's made without silicon, though it's obviously in the preliminary stages.
So essentially you want to get processors for little to no money, but must perform on the top of any other processor?
<--- China's that way, this is a capitalism country good luck with that.
DrMrLordX
Lifer
- Apr 27, 2000
- 22,547
- 12,413
- 136
I believe he's referring to a hypothetical 2 ghz dual-core CPU with gigs upon gigs of on-die cache.
A 2 ghz cpu with 2 gigs of 3-cycle L1 and 8 gigs of 12-cycle L2 certainly would be interesting, though I think yield problems would be so great that you would be better off developing sophisticated RAM systems that can run DDR2 or DDR3 so fast that you get latency of 50 cycles or less when running your processor in the 2-3 ghz range. Having RAM that fast would also make it easy for a system like Torrenza to allocate your system RAM for use by other add-on chips to use.
Granted, that would take some hella-fast RAM. I think DDR2-2400 5-5-5-12 2T would do the trick. Good luck with that.
DrMrLordX
Lifer
- Apr 27, 2000
- 22,547
- 12,413
- 136
There's been talk of transistors fabricated from materials other than silicon, such as diamond and other carbon-based materials. Nanotubes and nanotube-like structures have also been considered. Heat would be much less of a problem with those, I'd think.
Cache is certainly beneficial to a processor's performance, but the performance improvement afforded by RAM-sized caches would be nowhere near what you appear to expect. The reason for this is that typical caches (1MB+) have a hitrate in the region of 99.9%, which is difficult to improve upon. Hence diminishing returns for increasing cache sizes.
The point of cache is to improve a processor's performance by improving its efficiency (reducing wasted cycles). It is not a miracle solution by any means, and the transistors spent on such huge caches as those suggested would be better spent on core improvements, or more cores.
Well, going beyond skin deep for a moment there is more to multicore race then number of cores just like there was more to the GHz race then the frequencies (IPC).
Multicore design can go for quite a few different ways, we got:
[*]Homorganic Vs heterogenic.
[*]General purpose Vs specialized.
[*]Many simple cores Vs few big ones.
[*]Virtualization in CPU design* Vs "every core for itself".
* Hyperscalar (instead of superscalar), The final multicore CPU can be presented to the bios/OS as one CPU but inside it, it is composed of many cores.
Multicore design can go for quite a few different ways, we got:
[*]Homorganic Vs heterogenic.
[*]General purpose Vs specialized.
[*]Many simple cores Vs few big ones.
[*]Virtualization in CPU design* Vs "every core for itself".
* Hyperscalar (instead of superscalar), The final multicore CPU can be presented to the bios/OS as one CPU but inside it, it is composed of many cores.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 24K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 20K
-
-
-
Facebook
Twitter