Differences between Intel QPI and HT used by AMD

[ddkput]

I have been on a lot of sites and blogs and researched WIKI and without getting into a discussion on what better I would like to know what are the differences between QPI and HT. I have read that HT may be overall faster if the technology was there to utilize its full potential, but for now QPI offers less bottle necks and lower latencies with comparing apples to apples. Now I know to do a true comparison all things need to be equal. You know as well as I do that is truly impossible because of motherboard options and hundreds of CPU out there. Just in general with out needing an engineering degree to understand it, are there any major differences between the way they both transmit and receive info. I have looked at diagrams of QPI and HT in action but to many people argue back and forth about the presence of bios toward one or the other. So I do not know how accurate these are because of Product loyalty.

 

[ddkput]

Its hard to believe that no one has a comment or an answer for this. Usually I get more info than I need . whts up with that?

 

[truegenius]

i experimented my ht from 0.2ghz to 3.5ghz (stock is 2ghz) but i found no to very little difference in my games, apps, benches etc. I only have amd build thus can't test qpi/fsb.

So i don't think that this internal communication matters much, it will matter in server segment.

 

[hunter315]

Bus speed generally isn't the limiting factor, a decent implementation of either will make the difference between QPI and HT a moot point since neither one will be the slowest part. Just like on GPUs the width of the memory bus only makes a difference on things that have intense memory usage, with the HT and QPI buses it would be for something that needed massive amounts of data flowing in, maybe if you were doing incredibly simplistic math on a massive data set stored on a couple of those PCI-E SSDs in parallel you might saturate the bus but again its a very specific application.


The top of the line hardware is not meant for a system of the level that most of us will ever touch, QPI and HT are both meant for dual and quad CPU servers doing massive number crunching. Basic office stuff, gaming, and even most rendering systems don't have the necessary number crunching power to be able to saturate the bus so its really a moot point, they are different ways to get the same results.

 

[jimmysmitty]

What hunter said. QPI and HTT were ideas for servers that made their way to desktop. They are essentially overkill for most DT applications. FSB was good for a time but recently (multicore CPUs) it would show its age.

As for the difference? There is just a lot of little things. They essentially do the same thing, QPI and HTT. They both connect the CPUs, RAM and SB to eachother.

But one thing is that QPI is no longer in DT CPUs, only server CPUs. LGA2011 has it as a inter CPU connection but the rest of the system is using DMI, which was present when Intel was still using the FSB only its much faster.

All Intel DT CPUs use DMI to connect the CPU to the SB, PCIe, RAM etc and still its overkill. My 2500K has memory bandwidth of about 21-22GB/s with my current memory setup.

 

[ddkput]

Best answer selected by ddkput.

 

[ddkput]

Bus speed generally isn't the limiting factor, a decent implementation of either will make the difference between QPI and HT a moot point since neither one will be the slowest part. Just like on GPUs the width of the memory bus only makes a difference on things that have intense memory usage, with the HT and QPI buses it would be for something that needed massive amounts of data flowing in, maybe if you were doing incredibly simplistic math on a massive data set stored on a couple of those PCI-E SSDs in parallel you might saturate the bus but again its a very specific application.


The top of the line hardware is not meant for a system of the level that most of us will ever touch, QPI and HT are both meant for dual and quad CPU servers doing massive number crunching. Basic office stuff, gaming, and even most rendering systems don't have the necessary number crunching power to be able to saturate the bus so its really a moot point, they are different ways to get the same results.

Thanks for the info. I pretty much had about the same thing, but you did give me a little more insight.

 

[ddkput]

What hunter said. QPI and HTT were ideas for servers that made their way to desktop. They are essentially overkill for most DT applications. FSB was good for a time but recently (multicore CPUs) it would show its age.

As for the difference? There is just a lot of little things. They essentially do the same thing, QPI and HTT. They both connect the CPUs, RAM and SB to eachother.

But one thing is that QPI is no longer in DT CPUs, only server CPUs. LGA2011 has it as a inter CPU connection but the rest of the system is using DMI, which was present when Intel was still using the FSB only its much faster.

All Intel DT CPUs use DMI to connect the CPU to the SB, PCIe, RAM etc and still its overkill. My 2500K has memory bandwidth of about 21-22GB/s with my current memory setup.

I am not sure we are 100% on the same path, Pretty much though. HT to me is Hypertransport.... HTT I thought was Hyper Threading Technology. Anyway were were right with your recommendation about hunters statement and I also knew what you meant. I found a forum thread on overclockers.com that had a lot of praise and i guess it had some diagrams too. to old of a post for me to pull up all the old info. Sounds like the people who read it or ask for the info really liked the answer and understood it more than other sites postings, maybe the responder spelled it out in laymans terms and with the diagrams that just was the icing. Again thanks.

 

[jimmysmitty]

I am not sure we are 100% on the same path, Pretty much though. HT to me is Hypertransport.... HTT I thought was Hyper Threading Technology. Anyway were were right with your recommendation about hunters statement and I also knew what you meant. I found a forum thread on overclockers.com that had a lot of praise and i guess it had some diagrams too. to old of a post for me to pull up all the old info. Sounds like the people who read it or ask for the info really liked the answer and understood it more than other sites postings, maybe the responder spelled it out in laymans terms and with the diagrams that just was the icing. Again thanks.

Ahh. I see HT and think Hyper Threading, I se HTT and think HypreTransport Technology.

Still good to see you got your answer.

 

[Mousemonkey]

[:thegreatgrapeape:7]