Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

High End CPUs – Intel vs. AMD

cpubenchmark.net

Post reply on HN
Print view View original

41–50 of 98 posts

Re: High End CPUs – Intel vs. AMD

#41

Post by thethirdone »

siberianbear wrote: post #24
thethirdone wrote: post #21

Earlier quoted context omitted.

I get that, but given that GPUs currently can't use (for physical connector) / don't benefit from more than 16 lanes. Unless you are using multiple GPUs or another PCI device, you would not benefit from more than 16 lanes. How does that hold GPU manufacturers back?

Some parallel computing problems are best solved by one CPU and a lot of GPUs. In some of these problems, inter-GPU communication is important. If you have each GPU using a single PCI lane, then the compute power can be bound by iner-GPU communication bandwidth.

Yeah, there are definitely cases for threadripper's PCI lanes. But a single GPU is not one of those cases.

Re: High End CPUs – Intel vs. AMD

#42

Post by 0xbear »

In case someone from AMD is reading this: guys, you need to fix stability issues on Linux, or your Epyc is DOA. People are having some serious lock-up trouble with Ryzen, even with the latest AGESA updates and kernels. This is the only issue preventing me from recommending to purchase threadripper workstations at work. We do need that pcie bandwidth for GPU, but we absolutely can't tolerate instability.

Re: High End CPUs – Intel vs. AMD

#43

Post by foepys »

0xbear wrote: post #42

In case someone from AMD is reading this: guys, you need to fix stability issues on Linux, or your Epyc is DOA. People are having some serious lock-up trouble with Ryzen, even with the latest AGESA updates and kernels. This is the only issue preventing me from recommending to purchase threadripper workstations at work. We do need that pcie bandwidth for GPU, but we absolutely can't tolerate instability.

Epyc has another stepping than Ryzen and, as far as I know, nobody could reproduce the "kill-ryzen" SEGFAULT on Epyc or Threadripper.

Re: High End CPUs – Intel vs. AMD

#44

Post by syshum »

ijidak wrote: post #38
slackingoff2017 wrote: post #35

Earlier quoted context omitted.

Truly brilliant for sure. Reminds me of Anders Hejlsberg who designed Typescript, Delphi, and C#. Or Linus Torvalds with Linux, and just to prove he wasn't a one hit wonder designing Git. I've concluded that what people fail to grasp about technology is that Einstein types still exist. There are rare engineers worth hundreds of times more than average, only this time they're well paid by their corporate masters. More…

Throw in Jobs. Original Mac (1984), iPod (iconic music players), iPhone (nuf said), iPad (launch of tablets). Musk, Edison, Bell. Really a lot of the world changing in tech is concentrated in bursts of productivity for a relative few individuals like Einstein's golden year. Of course many make significant one off contributions. But it is amazing how some just have this ability to strike gold consistently. Making it l…

Jobs was a Marketer of other peoples ideas... I love how you leave out Woz, and the Team of people that made the iPod/iPhone/iPad

Edison is renowned for taking credit for other peoples work, notably some of the work of Nikola Tesla who you also leave off your list and was one of the rare singular geniuses on the level of a Einstein

Elon Musk has a vision of his idealized world and the money to hire people that will attempt to make it happen

None of the people you list (aside from Einstein) are singular genius working on their own creating the ideas and products and I think we do a disservice to everyone when we say "Jobs invented the iPhone".

That is not to say their contributions to the projects are not important but I do believe there is some amount of Hero Worship and over inflation of their value/importance and an massive under appreciation for the 100's or 1000's of other people that actually make these ideas or products a reality

Re: High End CPUs – Intel vs. AMD

#45

Post by pjmlp »

slackingoff2017 wrote: post #35
axaxs wrote: post #5

I don't mean to diminish the efforts of anyone involved, but I truly feel one man more or less moves the direction of the CPU industry: Jim Keller. He, among others, invented x86_64 at AMD during its previous glory days. AMD dominated the competition. He led the Apple chips at A4, and Apple chips then and now dominate the mobile competition. He came back to AMD and helped create Zen, to obvious results. Apparently he…

Truly brilliant for sure. Reminds me of Anders Hejlsberg who designed Typescript, Delphi, and C#. Or Linus Torvalds with Linux, and just to prove he wasn't a one hit wonder designing Git. I've concluded that what people fail to grasp about technology is that Einstein types still exist. There are rare engineers worth hundreds of times more than average, only this time they're well paid by their corporate masters. More…

Git was helped by Github, and being a requirement for Linux development.

Otherwise I might be enjoying Mercurial nowadays.

Re: High End CPUs – Intel vs. AMD

#46

Post by daemin »

dis-sys wrote: post #30
thethirdone wrote: post #21

Earlier quoted context omitted.

I get that, but given that GPUs currently can't use (for physical connector) / don't benefit from more than 16 lanes. Unless you are using multiple GPUs or another PCI device, you would not benefit from more than 16 lanes. How does that hold GPU manufacturers back?

16 lanes is required unless you are ready to give up 20% gaming performance on latest games. see the side by side comparison below. https://www.youtube.com/watch?time_continue=2&v=8KOTiee5RqA on top of that, many people want to stream their game video when they play their games on full speed, then a couple NVMe SSDs eat those lanes as well.

But if you're dropping 50% of the lanes and only getting a 20% speed drop, then it looks like you're not using those 16 lanes to capacity. Logically you'd expect closer to a 50% performance drop.

Re: High End CPUs – Intel vs. AMD

#47

Post by 0xbear »

foepys wrote: post #43
0xbear wrote: post #42

In case someone from AMD is reading this: guys, you need to fix stability issues on Linux, or your Epyc is DOA. People are having some serious lock-up trouble with Ryzen, even with the latest AGESA updates and kernels. This is the only issue preventing me from recommending to purchase threadripper workstations at work. We do need that pcie bandwidth for GPU, but we absolutely can't tolerate instability.

Epyc has another stepping than Ryzen and, as far as I know, nobody could reproduce the "kill-ryzen" SEGFAULT on Epyc or Threadripper.

I sure hope so, but given the epic fail of Ryzen, I'm going to wait for others to beta test it for me, and/or for some sort of explanation and fix from AMD.

Re: High End CPUs – Intel vs. AMD

#48

Post by dis-sys »

daemin wrote: post #46
dis-sys wrote: post #30

Earlier quoted context omitted.

16 lanes is required unless you are ready to give up 20% gaming performance on latest games. see the side by side comparison below. https://www.youtube.com/watch?time_continue=2&v=8KOTiee5RqA on top of that, many people want to stream their game video when they play their games on full speed, then a couple NVMe SSDs eat those lanes as well.

But if you're dropping 50% of the lanes and only getting a 20% speed drop, then it looks like you're not using those 16 lanes to capacity. Logically you'd expect closer to a 50% performance drop.

each Intel upgrade gives you 5-10% performance improvement. 20% is like 3 generations of Intel "innovation".

Re: High End CPUs – Intel vs. AMD

#49

Post by josteink »

thethirdone wrote: post #21
MattSteelblade wrote: post #16

Earlier quoted context omitted.

The vast majority of Intel's lineup--at least for the last few years--is only 16 lanes.

I get that, but given that GPUs currently can't use (for physical connector) / don't benefit from more than 16 lanes. Unless you are using multiple GPUs or another PCI device, you would not benefit from more than 16 lanes. How does that hold GPU manufacturers back?

> Unless you are using multiple GPUs or another PCI device

You mean like having a GPU and one or more PCIE NVMe storage-devices?

Post reply on HN