Originally by: Savvy

---

So I was wondering if Eve was going to be coded to make use of multiple CPu's on the client machine or if it was going to be optomised for 64 bit only? It does seem that both Intel and Amd are looking at multiple cpu options now to resolve the speed issues.

Also is eve going to be coded to allow for out of sequence code execution to make better use of the multicore cpu's we are seeing available now.
It seems pointless to have multiple cores if the cores are waiting for a thread to execute on another core before they can perform their task.

And Lastly are the develoopers looking at the physics engine cards as a way to speed up processing on the client side? an example being http://uk.asus.com/products1.aspx?l1=19

Ah back to Amiga days 1 chip for processing 1 for graphics and 1 for physics :)

---

Originally by: Xianthar

---

Originally by: Savvy

---

So I was wondering if Eve was going to be coded to make use of multiple CPu's on the client machine or if it was going to be optomised for 64 bit only? It does seem that both Intel and Amd are looking at multiple cpu options now to resolve the speed issues.

Also is eve going to be coded to allow for out of sequence code execution to make better use of the multicore cpu's we are seeing available now.
It seems pointless to have multiple cores if the cores are waiting for a thread to execute on another core before they can perform their task.

And Lastly are the develoopers looking at the physics engine cards as a way to speed up processing on the client side? an example being http://uk.asus.com/products1.aspx?l1=19

Ah back to Amiga days 1 chip for processing 1 for graphics and 1 for physics :)

---

optimizing the client for 64 bit will give no benefit to performance that i can see....what 64 bit math happens on the client? does the client use more than 4GB of memory? so why does 64bit computing matter to the client? it doesn't...ignore the hype :P as Intel put is best (paraphrased) "we had to develop a 64-bit processor to develop our new 32-bit processors with"....it will still be years before any MMPOG client makes real use of a 64-bit proc.

multi procs would be nice on the client side but again the client is not the bottleneck at the moment...not to mention that "sequential" code is not built for multi procs, you have it backwards.

physics cards? the physics are done on the server...at that end they are better off pushing the time intensive algorithms into a FPGA. There are FPGA's made to fit into an opteron socket for use as a co-proc. they _may_ be able to do something interesting things with those, but only if the server is spending a larger amount of their time on 1or 2 particular algorithms...only CCP knows that for sure, if not they are better off adding nodes or dumping money into a genius who can figure out how to get a stackless python scheduler to work on multiple CPU's at a time without getting strapped down by the OS.

-xian

---