Register
Notebookcheck
, , , , , ,
search relation.
, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,
 

Post reply

Warning: this topic has not been posted in for at least 120 days.
Unless you're sure you want to reply, please consider starting a new topic.
Name:
Email:
Subject:
Message icon:

Verification:

shortcuts: hit alt+s to submit/post or alt+p to preview


Topic Summary

Posted by: pol233
« on: March 09, 2017, 15:09:34 »

Hello,
Facing customers claims over 2x4K monitors tb16 failures, and 12mbps only eGPU traffic with no internal display capacity, Dell answered officialy the VERY VERY UNFORTUNATE misdesign issue of x2 lanes wired on x4 AlpineRidge !
see http://www.dell.com/support/article/en/en/frdhs1/QNA44089/thunderbolt-3-data-transfer-speeds-up-to-40gbps?lang=EN ... posted 27,Feb! And the funny 2x [email protected] 30fps !!! 2.000$ + 500$ + 2x 1250$ to get only 30fps ... is'nt it beautifull ? Customer legitime claims are answered that it's normal, even while purchasing site previously said "data" (so without DP or hdmi traffic) transfert to 40Gbps ! Now, as Dell usually does, no more mention of that data traffic, but new PCIE x2 lane port ' correction since mid-february. "Early XPS adopters" untill today are very pleased to get for 2000 bucks that modern f#@!*ng device! thanks you DeLL, for ever !! PS: TDP difference between x2lane-only low power chip (that is NOT the chip on MB which is the FULL x4lane one) and normal TB3 chip is under 0.5W, banning us of running GTX970+ (1050ti incl) to play at home with these "smart" laptop. Note that 2-in-1, with very low CPU not gaming candidate, is x4 lane. Very good job, DeLL !
Posted by: AquaeAtrae
« on: February 08, 2017, 20:57:29 »

x2 PCIe lanes ≠ 20 GBps ≠ 50% FPS in games

I'm curious to read the original report where a member saw 50% performance loss on x2 lanes. Thunderbolt 3 and eGPUs aren't the most intuitive technologies and I suspect there may be some confusion.

If you look at PCIe scaling tests (direct link below), it suggests that framerates in games actually aren't heavily affected by lower PCIe bandwidth. Also, the technical brief helps clarify how x2 or x4 lanes of PCIe bandwidth is actually passed bidirectionally through Thunderbolt 3, combined and prioritized with other signals.

Technical Brief
https://thunderbolttechnology.net/sites/default/files/HBD16235_Thunderbolt_TB_r05.pdf

PCIe Scaling Tests for Gaming
https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080_PCI_Express_Scaling/24.html

Note: x2 PCIe (gen 3) was not tested but has the same bandwidth as x4 PCIe (gen 2) which was tested. The later showed just a small loss (7%) in FPS compared to having double the bandwidth. This suggests that the actual impact in games is much less than the 50% so many fear. In other words, even today's most powerful GPUs (GTX 1080) doesn't fully utilize more than x2 PCIe lanes very often. It may occur occasionally as data is loaded, but overall gameplay is only slightly affected and more than made up for with the power of a desktop GPU.
Posted by: SpeeDeemon
« on: January 09, 2017, 00:15:35 »

Claimed 40GB/s and actual 16GB/s don't add up to "clearing Dell of wrongdoing" from where I'm sitting - quite the opposite, actually, no matter what technicalities from Intel may be involved.
Posted by: menpasav
« on: January 08, 2017, 18:57:21 »

Dell could choose between 2 solutions and follow TB3 spec -> why not.
BUT they shouldn't advertise a 40Gb/s data transfer compatibility -> this I believe is false information to consumers and one I used to chose this laptop (XPS13 9350).
Posted by: XCVG
« on: January 05, 2017, 08:51:37 »

As an unhappy XPS 15 owner, I wouldn't call this resolved.

Dell's marketing copy states that TB3 provides "data transfers up to 40Gbps" and compares it to USB3, implying that this is solely data throughput and does not include the DisplayPort streams. Even the 32Gbps actual bandwidth doesn't match that figure, but one could fudge that with rounding and overhead and probably get away with it. What we get- 16Gbps- is less than half of what we were led to believe we would get.

Within spec? Yes. What was advertised? No. Whatever technical justification they may or may not have does not justify swindling buyers out of half their promised speed.
Posted by: niemermer
« on: January 04, 2017, 19:07:53 »

Perfect! Someone needs a XPS 15? I have one to sell. And a big special thank to Dell and Intel!
Posted by: John Marsh
« on: January 04, 2017, 02:21:19 »

This is so disappointing. I've had my XPS 15 for 6 years now and ready to pull the trigger on a new one

Either the XPS 15 or a Macbook Pro with Soldered SSD. Don't know which is the lesser of 2 evils now.. I trusted Dell and you guys must feel cheated
Posted by: Markus Krainz
« on: January 04, 2017, 01:08:25 »

PS: I really would not use the phrase "added information on TB3 from Intel clearing Dell of wrongdoing".  Dell writes about Thunderbolt 3 all the time "Supports faster data transfer of up to 40 Gbps, ideal for large files." So while the Intel Alpine Ridge controller technically supports being driven by just 2x PCIe3.0 lanes, it makes no sense for Dell as that will cap file transfer at 16Gbit/s (- protocoll overhead), instead of 32Gbit/s + some reserve for Displayport (- protocoll overhead). I imagine this has to be an oversight on their part. It makes no sense to not connect the unused PCIe lanes from the CPU. And there might be enough substance there to get legal claims going for people who purchased it. Because clearly a potential buyer would not expect the laptop to be capped like this. Disclaimer: IANAL. And I am more interrested if the refreshed models will get a fix.
Posted by: Markus Krainz
« on: January 03, 2017, 19:48:51 »

Chance that the 2017 refreshes due this month at CES are affected?
Can we make Dell aware of the problem so at least the updated models are wired correctly?
Posted by: Adriano
« on: January 03, 2017, 12:44:17 »

Actually I do believe there is wrongdoing from dell: they marketed the port as 40Gbps while it simply is not. Isn't this true?
Posted by: Douglas Black
« on: January 03, 2017, 07:54:34 »

Thanks for the clarification, Nando. Will update.
Posted by: nando4
« on: January 03, 2017, 07:37:12 »

Douglas, a further update with reference to Intel Thunderbolt3 documentation clears Dell of any wrongdoing:  https://egpu.io/forums/pc-setup/xps-15-9550-not-able-to-reach-40gbps-over-tb3/paged/3/#post-980
Posted by: nando4
« on: January 03, 2017, 06:54:16 »

Douglas, this article is inaccurately referring to the 40Gbps TB3 channel as the problem. The issue is the notebook's PCIe link to the TB3 controller. There are only 2, rather than 4 PCIe lanes wired to the TB3 controller as specced by Intel. Pls refer to https://egpu.io/forums/pc-setup/xps-15-9550-not-able-to-reach-40gbps-over-tb3/paged/3/#post-979
Posted by: Redaktion
« on: January 03, 2017, 04:29:56 »

Schematics of the XPS 9350 show that its TB3 controller is hardware-limited to speeds half of those advertised. This design flaw may affect the XPS 9360, 9550, 9560, M5510, and M5520.

http://www.notebookcheck.net/Design-flaw-in-XPS-9350-9550-may-cause-TB3-ports-to-run-at-TB2-speeds.189743.0.html

 
C 2018 » Impressum     Sprachen: Deutsch | English | Español | Français | Italiano | Nederlands | Polski | Português | Русский | Türkçe | Svenska