Processors
Intel® Processors, Tools, and Utilities
14403 Discussions

Intel Xeon hpc solutions how do we know

idata
Employee
2,122 Views

I

have a big question, one i think that may have played on many minds. It sure is on mine and its left me wondering in which direction to go what investment to buy etc.

Its fair to say that the market of the core i9 has bled over into the Xeon market not surprising as people want more and more power grrr power yes power lol.

Only joking. What my problem is is that all the advertising efforts and product data that Intel have released seems to revolve one around windows, two around graphics, three around virtual machine servers.

There is very little data concerning my line of work, performance data regarding HPC for linux environments, compile times, linux graphics workstation performance, Reliability and development data for person like me who spends my life in low level hex assembly and C code.

Presently Large code compiles in linux can take hours, not days luckily. Compiling while at the same time as working on further more code, working with graphics front end, playing music(which never seems to be a problem), playing a video flash or other content never the less seems to be a major concern of mine with these processors.

My current dual xeon platform when compiling using the multi-thread power of each processor still takes hours but slows down and bogs the graphics performance(some fault maybe ATI's mind you). It also bogs down video decoding/encoding generally every form of computer task. With Intel new hardware what which processors can actively manage and handle high end multitasking situations like this not the above is just a few examples so many other things may be happening aswel.

It would seem to me Intel have designed processors to manage these tasks well in there new line of chips but when graphics design is not your main task, when virtual OS not of much interested apart from testing a kernel or code etc, when games are not even in the equation what would make my best solution from intel.

7500 is on the nahalem ex arch now which seems interesting in my solution, the 6500 that has recently popped up looks good. I can assign half the threads to compile while using the rest for other work from what i can gather. Knowing my luck i will buy it and half the threads or not it will still bog it down somehow lol.

Any help is appreciated even if you can point me in the right direction to research and make sure I can get the hardware that solves my problem.

I seriously beleive Intel has the solution just need to digg it out somehow.

Thanks again.

0 Kudos
4 Replies
idata
Employee
486 Views

I took a look at the benchmarks. Well they look good but how the hell am i supposed to turn that a reality. Generally a mainboard float around £600 gbp to house the processors. There is no boards to use these processors in the solution i am looking for. I have looked at making my own board seems possible but i noticed the chipset used for this series of processors does not have the PCIe interface i need. Its a gen2 x8 leaving me wondering.

I cant find any other chipsets compatible with this series of processors at present and i also don't think any board providers will be manufacturing boards to fit my solution.

Placing a renderfarm at the side of the machine seems an extreme aproach seen as intel's latest technology can be compacted inside one box.

0 Kudos
Edward_Z_Intel
Employee
486 Views

What's wrong with PCIe Gen2 x8 slots? They should be backward compatible with PCIe Gen1 cards...

0 Kudos
idata
Employee
486 Views

Good question. Well as far as using today hardware or current hardware is concerned for a consumer its fine. The problem i have is when i am working on my projects especially on some fpga designs i am trying to utilize the pcie gen2 x16 links through to gen3.There are ways around it but i would really like to be able to develop and test it all in one place.

With gen 2 x8 id be limited to testing and building it to the slower speed. If it was x16 i could live with that as i don't think gen 3 would be something to embed in the next years or at least just yet anyway.

I could do the code on this platform then send it over to a second machine which would have the interface but if i am going to build a machine for purposes like this why not build it with everything in one place from the start.

Considering the price and effort in building a system like this it has to fit the bill so to speak.

x8 just seems slow for a machine of this type. I can understand intel focusing on IO for a server/blade. Superfast system yet if you cant IO as fast eek.

0 Kudos
Reply