10GB ethernet Card?

The results are in... The ATTO drivers work on the ADD-ON card.

The Model NS12/NT12 were used, and successful, as this card has 2 RJ-45 ports on it.

When using the ATTO drivers, the card will take on the full appearance of being ATTO.

But, I have adjusted the ATTO INF file(s) to read the actual given card name, as seen here:


INF file for using ATTO drivers on this Intel ADD-ON card Windows XP 64-bit.

Windows XP 64-bit version of FFRM.inf (presently in use)

UNTESTED AT TIME OF POSTING
Windows XP 32 bit version of FFRM.inf modified for the ADD-ON card:

To unlock their extra functions, though they haven't been extensively tested, Lenovo had a driver package (tho no drivers registered as being compatible depsite listing XP 32/64) will enable their extra functions in adapter configuration:

Lenovo Intel tq1etn12sr17.exe (THE BOLD part being the original file name)

Found here
https://support.lenovo.com/us/en/do...r-2003-r2-thinkserver-rd330-rd430-rd530-rd630

Hope this helps!

 

After testing, the modified drivers with Windows XP 32-bit, it works. However, the ADD-ON adapter seems to only operate at 1.4 gbps on 32-bit XP. Odd.

I did not test the ATTO adapter on Windows XP 32-bit before switching to this card. I suspect though, it will be the same where it comes in with a 1.4 gig connection.


So, I've got two good working cards. A dual, and a single. Leaving the dual card from ADD-ON in.

 

I have scoured web searches dozens of times over after posting my findings, but still cannot find any OTHER compatible drivers than what ATTO has offered on their website. This is only taking official websites into consideration.

Now, some drivers do give the appearance of being compatible, but they report back a "Code 10" Error, which means the driver isn't actually compatible and the adapter won't work with those drivers.

As far as I'm concerned, the ATTO drivers should be able to work with any card that is the equivalent to Intel X520 or X540 T1 or T2 (the T1/T2 stands for RJ-45 ports/twisted pair number of plugs).

If you don't want to use the modified ATTO INF file, they will work with the original but will display with an adapter name of ATTO FastFrame.

The Lenovo package does appear to unlock additional things, but doesn't provide a useful driver for XP.

 

How does this product measure up?

Brocade ICX6450-24P
https://www.ruckussecurity.com/ICX-6450-24P.asp

https://www.ebay.com/sch/i.html?_fr....m570.l1313&_nkw=Brocade+ICX6450-24P&_sacat=0


Would you consider it a cheaper but equivalent version of the Cisco 350X's?

 

Here is an ISO containing PROXGB drivers.

https://www.ibm.com/support/pages/d...l-based-ethernet-ibm-bladecenter-and-system-x

It's version 15.1 of Intel's Network adapter drivers. The INF contained in there for v5.2 of windows predates the drivers from the ATTO driver pack.

Also, there is also version 15.6, which seems to be dated after the ATTO driver pack.

https://www.ibm.com/support/pages/v...l-based-ethernet-ibm-bladecenter-and-system-x


When I have the ability to do so, I'm gonna do some tinkering and figure out if I can directly use the drivers from either pack for XP64 (and later XP32).


It could very well turnout that nothing but the ATTO drivers will work.

 

None of the ISO files found on IBM's website relating to Intel PROXGB drivers worked.

They would install, but after a message saying that windows cannot insure the drivers work. The drivers do install, but the hardware/device registered a "Code 10" which translates as being incompatible hardware/driver.

So, for the time being, ONLY the ATTO drivers actually work.

I went back as far as driver packs version 15.1, but nothing worked.

I am going to see if maybe a combination of files from ATTO and driver packs can work some sort of way. But, anyway, 10G ethernet cards can work, especially on XP 64-bit.

 

I have discovered that if I used the INF file I compiled from ATTO INF file, while adding other adapters to the list, along with using whatever driver files that exist in NDIS5x folders, you can effectively use the drivers from any driver pack.

I also discovered that there's some differences between the INF as found in the IBM Intel 17.4 PROXGB driver ISO and others.

I plan to work towards coming up with a "perfected" INF file to use with the newer drivers. It may work towards expanding XP 32 and 64 bit systems using practically any 10G network card.

I will make an update to this thread with my finds and comentary, and the other in the hardware section with the "perfected" INF files.

 

I was able to do a INF "merge" from the ATTO drivers to a more recent (later) version of the Intel drivers as released on the version 17.4 release.


v17.4 PROXGB driver ISO file.
https://www.ibm.com/support/pages/node/835632

As far as these drivers actually working... well, my XP 64 is working just fine with an X540-T2 adapter, even though I selected the name "X520-T2" (though really not much difference between the cards).

It's entirely possible that the INF files created can be used with just about any driver pack that shows support for Server 2003 and/or has NDIS5x files.

 

I have spotted a later driver on IBM that supports "2003" Windows, as denoted by its contents containing NDIS5x folder.

https://www.ibm.com/support/pages/i...icrosoft-windows-ibm-bladecenter-and-system-x

This appears to be the latest pack supporting it. Other v18.7 pack (an ISO found on IBM website) does not have it present.

 

Very interesting work!

I ran across this type of driver issue trying to find a native xp32 driver for a perc5i raid controller for a Dell Poweredge 2950 server that I wanted to use with xp. Took a long time to find the driver after doing basically what you are of finding every single driver in existence. Don't give up! You'll find it! And then let us all know please.

 

Bad news. It seems as though any card, even with the ATTO drivers, will never pull anymore than 10% of it's 10gigabit bandwidth. So, effectively, the cards operate as 1 gigabit.

From directly to the NAS from my computer, I pulled in at most 1 gigabit between both devices.

Then, I installed the ATTO card on a second machine and used the X540 card I have to directly attach both PCs together. Swapping files via a SMB explorer share, going from one station to the other: SSD to ram disk. The transport speed never exceeded 10% of the 10Gigabit.

While off the network still, and directly attached to the other XP64 machine and the NAS, I copied to both devices at the same time. The X540 card has two ports on it. Between the two, the card was still working at 10% capacity, this time 6% for one transfer and 4% for the other.


So, it would seem that the 10giga cards, albeit working, appear to be working at 10% capacity. The cards and the switches all report back as being 10gig. Never seeming to get above 1gigabit transfers.

Unless I'm mistaken, this appears to be a bust.

 

I have rang Jen IE Mr Siamese Cat; And asked him to drop you data for our tests - I didn’t take any notes.

 

Madeleine has asked me to take a look at this thread. I think she has forgotten the times recorded while experimenting when we tested 1Gbit unmanaged single cable up against 2Gbit Managed twin cable.

Both Madeleine and myself use solid state, in Madeleine’s case M2. All are either Raid 6, Raid 5 or Raid 0. Neither of us use Raid 1 on anything, it is far too slow. We both also use SCSI, but only tested Madeleine’s as mine was too heavy to take over. Madeleine’s 14 Drive Raid 6 SCSI was slightly slower than M2 Raid 5 solid state.

So lets make it real simple and just give you the best results (Real World) solid state Raid 0 to solid state Raid 0 both on Supermicro’s across the network. This way the disks are not causing the bottleneck or so we hoped. (More on this at the end)

It is really easy to get much faster results by using files that are not zipped or compressed or one large uncompressed file that they always use for the “Review” testing trying to sell you network cards or hard drives. I can easily get 75% of 1Gbit with small files. It is also easy to get high percentage network usage if you are throwing many different files simultaneously. We just wanted to keep it simple with less variables.

So we used various files some compressed, some large and stuck in a folder adding up to 10Gbytes.

10gigabyte (10,000Mb) folder with various files used. We stuck to this for simplicity of results. All timing was done with a stop watch. We weren’t interested in what the computers felt they were doing.

On my unmanaged 1Gbit system NetgearGS116 single cable there was a massive variation, but the average was 7 minutes 15 seconds. 435 seconds (Running at about 20% of 1Gbit)

Therefore Real World Network transfer speed unmanaged 1Gbit is 23Mb per second.

On Madeleine’s theoretical 2Gbit system twin cable managed Cisco SG300, there was far more consistent transfer rates, probably due to the managed switch alleviating the data crashing into data that I will have undoubtedly got.

Average was about 2 minutes 31 seconds. 145 Seconds (55% of 1Gbit)

Therefore Real World Network transfer speed Managed 2Gbit is 69Mb per second.

Apparently the theoretical transfer speed of a 1Gbit Network is 125Mbytes per second

And the theoretical transfer speed of a 10Gbit Network is 1250Mbytes per second.

Most review sites just talk utter twaddle regarding transfer speeds. For instance this well respected review site would try and have us believe that they achieved 112.5MB/s on a 1Gbit Network. That is almost the theoretical maximum. Hmm I wonder what special file they used for that test?

https://www.techspot.com/review/528-nas-10gbe-performance/page5.html

And now for the punch line.

I used to have two Raid 0 disks on my system made up of 4 EVO 860 500GB

Transferring the same 10GB folder from Raid 0 to a different Raid 0 on the same system took about 2 minutes.

This gave me a Real World transfer speed of 83MB/s

But apparently according to SiSoftware Sandra my Raid 0 storms along at 407MB/s

And I think you will find this is the difference between adolescent boys on the internet trying to out-do each other and adults looking at actual real world speed of real world files tested by an actual stop watch.

Now my Supermicro is an older system running 32bit XP, so perhaps some people will do better. However Madeleine was using Windows 7 and 10 as well as XP.

I suggest you really need to get a stop watch out and see what speed you can transfer a 10Gbyte folder with say a few Zip files a couple of compressed video’s and an assortment of other files. You might find that it is your actual drives that are causing the bottleneck.

In any event even you upgrade to RAID 0 M2 all round, if you achieve a Real World Transfer rate of anything over 250MB/s (Just one fifth of theoretical 10Gbit speed) I will be very very impressed.

 

Very interesting. What slot are these cards in? Because if the pcie slot is limiting the bandwidth you can see this same issue, but then it's more of a motherboard/system issue versus a driver or nic one.

 

There is also another point to consider which is the possible limits of smb1. I know on my win7 systems where smb2 is enabled, nas transfer speeds can double. The coupling of smb2/3 and 10Gb is where the real speed increases seem to be, and xp may be limited in what it can do with a 10Gb link even when it has one.

 

No there isn't. You got proof?

 

Why have you copied and pasted a response I made to Secpar 18 months ago?


Really weird.


I will be flagging for Ian.

 

Spotted this today.
Listed as: Mugast E10G41BFSR/X520-SR1 PCI-E X8 Network Card,10Gbps 82599EN SFP+Ethernet Converged Network Adapter with Driver CD and Low Profile Bracket for Windows XP/VISTA/7/8/10, etc

https://www.amazon.com/Mugast-E10G41BFSR-X520-SR1-Ethernet-Converged/dp/B07YCJCVTF

The "82599" points it being Intel.

I just found it interesting that this is being sold as being for XP/Vista.

I wonder what drivers will be found on the Driver CD that's listed in the package list.