USB 3.0 Performance: Two Solutions From Asus And Gigabyte

[dhowie]

[citation][nom]Crashman[/nom]SSD's could probably do that now if they applied 24-way parallelism to the internal controller and the fastest available chips.[/citation]

+1

 

[bk420]

it's about time.

 

[bk420]

[citation][nom]Crashman[/nom]UBS operates in two locations, Everywhere, and right next to you.[/citation]
Universal Bus Cereal......hahahahhaahhahaha
j/k.

 

[bk420]

[citation][nom]liquidsnake718[/nom]BAH.... Im waiting for an X58 with USB.3.0 AND 16x 16x SLI. I would not want to sacrifice the other slot for a 8x config....[/citation]
add 16x *2 more...

 

[donkeykong2009]

This is a good review. I do hope that the reviewer can help 2 questions that came up based on their findings:
1. in a single card scenario, would the sacrifice down to pci-e running at x8 happen even if there are no USB 3.0 and SATA 3 devices connected
2. what is the real practical impact of running pcie at x8

I have a single 5870 ATI graphics, I was planning to buy the GA-P55A-UD4P, but now I am confused if its worth buying the significantly more expensive Asus. I have not had a very good experience with my last 2 Asus motherboards, they don't seem to be stable boards, good for overclocking but not good for stability (long running), so I am resisting to pay more for motherboards especially Asus ones. Also P55 is a dead end motherboard, so I expect to throw it away in a few years time, hence trying to cut costs down.

 

[Crashman]

[citation][nom]donkeykong2009[/nom]This is a good review. I do hope that the reviewer can help 2 questions that came up based on their findings:1. in a single card scenario, would the sacrifice down to pci-e running at x8 happen even if there are no USB 3.0 and SATA 3 devices connected2. what is the real practical impact of running pcie at x8I have a single 5870 ATI graphics, I was planning to buy the GA-P55A-UD4P, but now I am confused if its worth buying the significantly more expensive Asus. I have not had a very good experience with my last 2 Asus motherboards, they don't seem to be stable boards, good for overclocking but not good for stability (long running), so I am resisting to pay more for motherboards especially Asus ones. Also P55 is a dead end motherboard, so I expect to throw it away in a few years time, hence trying to cut costs down.[/citation]

If you disable the features in BIOS you get your eight lanes back for a total of 16. If you enable the features in BIOS you lose eight lanes.

 

[donkeykong2009]

[citation][nom]Crashman[/nom]If you disable the features in BIOS you get your eight lanes back for a total of 16. If you enable the features in BIOS you lose eight lanes.[/citation]

Thanks... going with Asus again ..

 

[Guest]

http://www.pricedizzy.com/processors.html I5 750 - $152.....................................................!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

 

[Crashman]

[citation][nom]zingam[/nom]Intel postponed USB 3.0, yesterday Larrabee. What's next? What are they going to fuck up now?[/citation]

Postponed? I'm not sure that's the proper way to put it. LGA-1156 is derived from LGA-1366, which was developed befor USB 3.0 was finished.

 

[impreza]

that could be really bad news if hp or dell stops other makes having front panel usb3. no front panel usb3 standard could pretty much kill usb3. you think you have your usb hdd but the plug on the front is only usb 2, most people wont even think that the one on the bakc could be way faster and just use the front since it is convenient.

 

[b82]

[citation][nom]scryer_360[/nom]Although its good to see the hardware in the wild, I don't yet see any reason to buy something for USB 3.0 and SATA 6GB/s.Yes there is some future proofing, but as I think it was mentioned at the beginning of the article, it will be years before we see a significant number of USB 3.0 and SATA 6GB/s devices on the market, and maybe even longer before we see storage with the ability to use all that bandwidth.Certainly it will take aeons for traditional hdds to utilize all that speed, but ssd's might be able to do that in a few years, right?[/citation]
Scryer, so its not worth adopting USB 3.0 until we have hardware with ability to use all of its bandwidth?? I don't think this is a very sensible argument. As soon as we can use just a part of USB 3.0 bandwidth we already have significant gains over USB 2.0, and that's progress. USB 3.0 may offer up to 10x more speed than USB 2.0, but if you're able to take advantage of just 4x USB 2.0, then when transferring several hundred GBs that is an advantage you will appreciate. Maxxing out an interface matters for testing, but all that matters for ordinary useage is does it deliver an improvement already over what came before? As one of the earliest graphs/charts in this review demonstrates, showing USB 2 way down the bottom, the answer is already a resounding YES. We don't need to wait for the interface to get maxxed out.

One other point, this review thoroughly covered the motherboards USB 3.0 performance. But the cost for the Gigabyte was cutting a graphics card down from PCIe 16x to 8x. I would like to have seem some anaylsis of whether the impact of this was major or minor and how much it depended on the graphics card and/or the graphics application being displayed. Without this the review seems incomplete.

 

[Crashman]

[citation][nom]b82[/nom]Gigabyte was cutting a graphics card down from PCIe 16x to 8x. I would like to have seem some anaylsis of whether the impact of this was major or minor and how much it depended on the graphics card and/or the graphics application being displayed. Without this the review seems incomplete.[/citation]

You're talking about more than one article's worth of data. First you must determine how much performance is lost going from 16x to 8x, there's an article for that:

http://www.tomshardware.com/reviews/p55-pci-express-scaling,2517.html

Of course you're asking for more cards to be tested. Knowing that you're looking at roughly 3 days of testing per card, how many would you like to add?

 

[b82]

[citation][nom]Crashman[/nom]First you must determine how much performance is lost going from 16x to 8x, there's an article for that:http://www.tomshardware.com/review [...] ,2517.htmlOf course you're asking for more cards to be tested. Knowing that you're looking at roughly 3 days of testing per card, how many would you like to add?[/citation]
Actually I would like to add no cards. Thank you for pointing out the article that discusses the 16x to 8x performance hit. That was usefull.

What I think the reviewer should have done is briefly mentioned that article with a hyperlink, just as you have done. That would have satisfied me.

 

[Crashman]

[citation][nom]b82[/nom]Actually I would like to add no cards. Thank you for pointing out the article that discusses the 16x to 8x performance hit. That was usefull.What I think the reviewer should have done is briefly mentioned that article with a hyperlink, just as you have done. That would have satisfied me.[/citation]

Well, to answer more completely as best I can (based on other articles), I believe the HD 5750 is the bottom-point where going from 16x to 8x no longer makes a consistent difference, so the difference for a 5770 would likely be measurable but not noticeable, etc.

 

[b82]

Obviously with socket 1156 there is no easy answer. Perhaps my perfect solution from Gigabyte would have been their existing one, PLUS the option in the bios to switch the USB 3.0 to one of the crippled 2.5 GT/s PCIe x4 lanes. While this would have meant USB 3.0 could not perform at its best, would it not have provided the user with the option to avoid impact on the PCIe x16 while still having a USB interface that was much faster than USB 2.0?

 

[b82]

[citation][nom]Crashman[/nom]If you disable the features in BIOS you get your eight lanes back for a total of 16. If you enable the features in BIOS you lose eight lanes.[/citation]
I really hope this answer is accurate. I am reading Gigabyte's manual for their new H57M-USB3.0 and their H55M-USB3.0 (available on their website). The manual's block diagram clearly shows USB 3.0 sharing bandwidth with the PCIe 16x. However, nowhere does the manual say that you have to disable the USB3.0 in the bios to get your full x16. Rather bad of them if the slot will actually run at x8 ALL the time USB 3.0 is enabled.