- Status
RADIO_ACTIVE
Distinguished
- Jan 17, 2008
- 897
- 0
- 18,990
[citation][nom]iqvl[/nom]Good news to peoples like me who haven't spent any money on PCIE 2.0 DX11 card[/citation]
Don't worry when you decide to buy next year DX 12 will be out lol! Then it will be good news to those who waited then and so on and so on and so on...
Don't worry when you decide to buy next year DX 12 will be out lol! Then it will be good news to those who waited then and so on and so on and so on...
I'd like to see manufacturers create graphics solutions that use USB 3.0 and then simplify motherboards by removing pcie and replacing it with more usb 3.0 ports. Making video cards with a usb 3.0 interface could enable laptops to finally have a much needed graphics boost.
Why do all people assume that PCI-e revisions are about graphics? This article is consume oriented, but PCI SIG has to move quickly because enterprise servers already saturate PCI-e 2.0 easily. RAID solutions top out at ~2800 MB/s on PCI 2.0, and let's not forget that SAS/FC/Ethernet are not sitting around waiting on the PCI SIG to make a new revision as they're working on theirs as well...
warmon6
Splendid
- Jul 24, 2009
- 4,169
- 5
- 22,965
[citation][nom]gallidorn[/nom]I'd like to see manufacturers create graphics solutions that use USB 3.0 and then simplify motherboards by removing pcie and replacing it with more usb 3.0 ports. Making video cards with a usb 3.0 interface could enable laptops to finally have a much needed graphics boost.[/citation]
Err..... no. USB 3.0 is not the way to go for laptop with graphics. The bandwidth would be to small. Unless im mistaken, USB 3.0 speeds are equal to PCI-e 2.0 x1 speeds. This would mean you would need to have 8 usb 3.0 ports on a single computer and the usb graphic dock just to have decent performance (equivalent to PCI-e 2.0 x8) without to much bottlenecking via bandwidth. Although with that many usb ports needing to be synchronized to just 1 thing, i would imagine that there would be some latency issues....
What we need is something like what gigabyte has already done with there M1405 notebooks. Have a dockable graphic station thats connect decent high powered graphics for gaming. Although someone need to make a station customizable with a gpu of your choice.
(also dont use bold letters. bold letters is the equivalent of shouting at someone 😉)
Anyways, Hovaucf has a point. Primary reason why pci-e 3.0 is comming around so soon is mainly the enterprise/sever market. There isn't really much on the consumer side that needs pci-e 3.0 yet. Even the gtx 480/HD5970 can just barely show the difference between the x8 and x16 bandwidth.
Err..... no. USB 3.0 is not the way to go for laptop with graphics. The bandwidth would be to small. Unless im mistaken, USB 3.0 speeds are equal to PCI-e 2.0 x1 speeds. This would mean you would need to have 8 usb 3.0 ports on a single computer and the usb graphic dock just to have decent performance (equivalent to PCI-e 2.0 x8) without to much bottlenecking via bandwidth. Although with that many usb ports needing to be synchronized to just 1 thing, i would imagine that there would be some latency issues....
What we need is something like what gigabyte has already done with there M1405 notebooks. Have a dockable graphic station thats connect decent high powered graphics for gaming. Although someone need to make a station customizable with a gpu of your choice.
(also dont use bold letters. bold letters is the equivalent of shouting at someone 😉)
Anyways, Hovaucf has a point. Primary reason why pci-e 3.0 is comming around so soon is mainly the enterprise/sever market. There isn't really much on the consumer side that needs pci-e 3.0 yet. Even the gtx 480/HD5970 can just barely show the difference between the x8 and x16 bandwidth.
This article misses the boat almost completely. While PCIe 3.0 doubles the throughput rate, it also lowers latency. That's where the breakthrough really is, as commenter ta152h notes above. CUDA apps will immediately benefit from lowered data transfer latencies, as will many other applications which currently are hobbled by system latencies.
Kind of reminds me of SSD reviews which always focus on data throughput and miss the latency story. Ask anyone who actually uses an SSD and they universally state that it feels like a whole 'nother computer because login, app startup and data input happen instantaneously.
Now combine PCIe 3.0 and SSD, perhaps with a PCI-based RAID controller driving several 6 GB/s SATA SSDs, or better still with flash right on the PCI card such as FusionIO's devices. Now we're talking dramatically enhanced IOPS performance. It's the latency, people.
Kind of reminds me of SSD reviews which always focus on data throughput and miss the latency story. Ask anyone who actually uses an SSD and they universally state that it feels like a whole 'nother computer because login, app startup and data input happen instantaneously.
Now combine PCIe 3.0 and SSD, perhaps with a PCI-based RAID controller driving several 6 GB/s SATA SSDs, or better still with flash right on the PCI card such as FusionIO's devices. Now we're talking dramatically enhanced IOPS performance. It's the latency, people.
This article misses the boat almost completely. While PCIe 3.0 doubles the throughput rate, it also lowers latency. That's where the breakthrough really is, as commenter ta152h notes above. CUDA apps will immediately benefit from lowered data transfer latencies, as will many other applications which currently are hobbled by system latencies.
Kind of reminds me of SSD reviews which always focus on data throughput and miss the latency story. Ask anyone who actually uses an SSD and they universally state that it feels like a whole 'nother computer because login, app startup and data input happen instantaneously.
Now combine PCIe 3.0 and SSD, perhaps with a PCI-based RAID controller driving several 6 GB/s SATA SSDs, or better still with flash right on the PCI card such as FusionIO's devices. Now we're talking dramatically enhanced IOPS performance. It's the latency, people.
Kind of reminds me of SSD reviews which always focus on data throughput and miss the latency story. Ask anyone who actually uses an SSD and they universally state that it feels like a whole 'nother computer because login, app startup and data input happen instantaneously.
Now combine PCIe 3.0 and SSD, perhaps with a PCI-based RAID controller driving several 6 GB/s SATA SSDs, or better still with flash right on the PCI card such as FusionIO's devices. Now we're talking dramatically enhanced IOPS performance. It's the latency, people.
Hupiscratch
Distinguished
- Oct 29, 2008
- 243
- 0
- 18,680
That´s why I said in the last motherboard article that I´d rather wait for the PCI-E 3.0. I was feeling this article coming.
razercultmember1
Distinguished
- Dec 28, 2009
- 439
- 0
- 18,780
[citation][nom]nforce4max[/nom]Oh my the pace of things these days. For all the clocks they could at the least improve I/O performance rather than just the signal clocks and bust bandwidth. The latencies are horrible with AIB (add in boards) and that fact hasn't changed since the ISA days. As for backward compatibility it better work as expected. I would be steamed if my pci-e 1.0 cards didn't work when I want to some retro.[/citation]
Really, you don't know what you're talking about.
The AT-BUS (incorrectly referred to as ISA) had very low latency, so much so that memory for the processor was put there, and was read in two clock cycles. Compare this to the desperately slow L1 cache of the Nehalem, which takes 4 clock cycles.
So, relatively speaking, you could read twice as fast from the AT-Bus memory card as you could from the L1 cache on the Nehalem, which, outside of the registers, is the fastest thing in the system.
Really, you don't know what you're talking about.
The AT-BUS (incorrectly referred to as ISA) had very low latency, so much so that memory for the processor was put there, and was read in two clock cycles. Compare this to the desperately slow L1 cache of the Nehalem, which takes 4 clock cycles.
So, relatively speaking, you could read twice as fast from the AT-Bus memory card as you could from the L1 cache on the Nehalem, which, outside of the registers, is the fastest thing in the system.
[citation][nom]gallidorn[/nom]I'd like to see manufacturers create graphics solutions that use USB 3.0 and then simplify motherboards by removing pcie and replacing it with more usb 3.0 ports. Making video cards with a usb 3.0 interface could enable laptops to finally have a much needed graphics boost.[/citation]
I'm pretty sure you're confusing things. How do you think USB 3.0 connects to the processor? A bus is what makes everything connect to each other, and without PCI Express, you'd need another bus to connect USB 3.0, which is kind of a sub-bus. The only way around that, which will not happen any time in the near future (I hate to say never) is adding USB 3.0 functionality to the processor, and adding pins and lines on the motherboard directly to the physical port. I'm sure you can see how absurd that would be.
The system bus has been around since DEC first used about half a century ago. I don't think there's a better way to do it. I'm sure you're not saying to use USB as the system bus. You'd have very, very poor performance.
I'm pretty sure you're confusing things. How do you think USB 3.0 connects to the processor? A bus is what makes everything connect to each other, and without PCI Express, you'd need another bus to connect USB 3.0, which is kind of a sub-bus. The only way around that, which will not happen any time in the near future (I hate to say never) is adding USB 3.0 functionality to the processor, and adding pins and lines on the motherboard directly to the physical port. I'm sure you can see how absurd that would be.
The system bus has been around since DEC first used about half a century ago. I don't think there's a better way to do it. I'm sure you're not saying to use USB as the system bus. You'd have very, very poor performance.
aneasytarget
Distinguished
- Aug 26, 2009
- 44
- 0
- 18,530
[citation][nom]pjmelect[/nom]Am I the only one to think the explanation of “not having enough zeros and ones” as the reason for backward compatibility problems as being a bit strange.[/citation]
It refers to them transferring from 8b/10b encoding to 128b/130b encoding.
It refers to them transferring from 8b/10b encoding to 128b/130b encoding.
Good, this time next year.. right? So that'll mean Socket 775 will see it's last breath too? Bah! DX 11 isn't seeing much of a requirement as there is still a myriad of standards going all the way back to 2003ish and it seems we're on a 6 - 10 year upgrade cycle thanks in no small part to the greed of major manufacturers: Intel, Amd, Nvidia, ATi(AMD again), Dell, HP, Microsoft, Apple[?!](yes, them too), Just about every Flash, Volatile Memory company and SSD manufacturer you can think of.. oh yeah.. and the WORLD ECONOMY, last but not least the OIL COMPANIES & Bush 2's policy 2000-2008 thrown in for good measure.
Nobody's jumping on the tech standards bandwagon that fast anymore.. pcie-3 isn't something somebody's just GOTTA have the way enthusiasts were even just 10 years ago, let alone the go-go days of the 70s, 80s, & 90s, and early 2000s. Maybe millions of idiots do that for Apple's products but that's about the only anomoly out there in tech, perhaps excepting for next generation broadband (20+megabits).
Nobody's jumping on the tech standards bandwagon that fast anymore.. pcie-3 isn't something somebody's just GOTTA have the way enthusiasts were even just 10 years ago, let alone the go-go days of the 70s, 80s, & 90s, and early 2000s. Maybe millions of idiots do that for Apple's products but that's about the only anomoly out there in tech, perhaps excepting for next generation broadband (20+megabits).
Im good with my Radeon HD 5970 2GB GDDR5 two GPU`s
It will probably last 5-7 years.
So i am fine
It will probably last 5-7 years.
So i am fine
liquidsnake718
Distinguished
- Jul 8, 2009
- 1,379
- 0
- 19,310
Oh great a X-68 motherboard was mentioned as well in this article..... I wonder if this also means more "common" and cheaper 6 core cpus coming out! just great right when i just bought a core i7 930! Oh well....
liquidsnake718
Distinguished
- Jul 8, 2009
- 1,379
- 0
- 19,310
[citation][nom]ta152h[/nom]I'm pretty sure you're confusing things. How do you think USB 3.0 connects to the processor? A bus is what makes everything connect to each other, and without PCI Express, you'd need another bus to connect USB 3.0, which is kind of a sub-bus. The only way around that, which will not happen any time in the near future (I hate to say never) is adding USB 3.0 functionality to the processor, and adding pins and lines on the motherboard directly to the physical port. I'm sure you can see how absurd that would be. The system bus has been around since DEC first used about half a century ago. I don't think there's a better way to do it. I'm sure you're not saying to use USB as the system bus. You'd have very, very poor performance.[/citation]
Well then one could argue that developers like Ati and Nvidia should focus on using onboard GPU's/integrated gpus and develope something to replace those pcie2.0 or 3.0 lanes for bandwidth...or connect these integrated gpus in a way that they can maximize the pcie bandwidth.
A usb 3.0 connection for a gpu is a good idea and im sure it has been thought of already.
Well then one could argue that developers like Ati and Nvidia should focus on using onboard GPU's/integrated gpus and develope something to replace those pcie2.0 or 3.0 lanes for bandwidth...or connect these integrated gpus in a way that they can maximize the pcie bandwidth.
A usb 3.0 connection for a gpu is a good idea and im sure it has been thought of already.
neiroatopelcc
Distinguished
- Oct 3, 2006
- 3,078
- 0
- 20,810
[citation][nom]Article[/nom]SATA 6 Gb/s (which no storage device can even come close to saturating). [/citation]
There's absolutely no reason to think a ddr3 based ram drive can't or that a san box won't emerge with esata support instead of iscsi or fiber channel.
There's absolutely no reason to think a ddr3 based ram drive can't or that a san box won't emerge with esata support instead of iscsi or fiber channel.
- Status
TRENDING THREADS
-
-
-
News AMD explains the missing RDNA 4 GPUs in its CES 2025 livestream
- Started by Admin
- Replies: 15
-
Question New build wont post Gigabyte B650M Gaming Plus wifi , AMD Ryzen 5 7600X CPU, 32GB T-Force RGB DDR5. Need help!
- Started by Notmiserable
- Replies: 27
-
Discussion PSU recommendations and power supply discussion thread - Tom's hardware- Started by Darkbreeze
- Replies: 2K
-
Discussion What's your favourite video game you've been playing?- Started by amdfangirl
- Replies: 4K
Facebook
Twitter
Instagram