Slashdot Mirror
Mixing Gigabit, Copper, and Linux
iampgray writes: "With copper-based gigabit cards selling for less than $36 these days, what kind of performance can you expect -- especially in the often-overlooked Linux market? We sought out to test exactly what you could expect from copper-based gigabit solutoins for the desktop interface through the cluster-targeted products. Name brands and off-brands were put through the wringer. How'd they fare? Interesting results to say the least."
Between 2 Tyan/AMB 1.2Ghz machines I was able to pull 800 mega bits per second on old copper.
Gigabit Over Copper Evaluation
DRAFT
Prepared by Anthony Betz and Paul Gray
April 2, 2002
University of Northern Iowa
Department of Computer Science
Cedar Falls, IA 50614
Given the relatively low cost, backwards-compatibility, and widely-availability solutions for gigabit over copper network interfaces, the migration to commodity gigabit networks has begun. Copper-based gigabit solutions are now providing an alternative to the often more expensive fiber-based network solutions that are typically integrated in high performance environments such as today's tightly-coupled cluster systems.
But how do these cards compare with their fiber based counterparts? Are the Linux-based drivers ready for prime-time? The intent of this paper is to provide an extensive comparison of the various Gigabit over copper network interface cards available. Since performance is based on numerous factors such as bus architecture and the network protocol being used, these are the two main subjects of our investigation.
Our bandwidth benchmarks look at sustained throughput using TCP. While other communication protocols are available, indeed preferred, for high-performance computing, TCP-based benchmarks provide an immediate insight into the expected performance of the cards. With PCI-X coming into the marketplace in more and more motherboards as well as the multitude of systems with more traditional 32-bit PCI subsystems, numerous cards are available for today's 64bit and 32bit computer systems. The 64bit cards tested were as follows: Syskonnect SK9821, Syskonnect SK9D21, Asante Giganix, Ark Soho-GA2000T, 3Com 3c996BT and Intel's E1000 XT. The 32bit cards were Ark Soho-GA2500T, D-Link DGE500T. Comparisons for the various cards were made with respect to operation in alternate bus configurations and varied maximum transmission unit (MTU) sizes of TCP frames (jumbo frames). Results were gathered using Netpipe 2.4. By using Netpipe the peak sustained throughput would be provided as well as the transfer rate for varying packet sizes.
Note: All cards were tested at 1500, 3000, 4000, and 6000 values for the TCP MTU size. The drivers for the cards were not modified. Cards based upon the dp83820 chipset were limited to 6000MTU due to driver defaults. All other cards were tested through 9000MTU.
Cards Tested and Document Links:
D-Link DGE 500T (32-bit)
ARK Soho-GA2500T (32-bit)
ARK Soho-GA2000T
Asante Giganix
Syskonnect SK9821
Syskonnect SK9D2
3Com 3c996BT
Intel Pro 1000 XT
Comparisons and Observations
Our Testing Environments:
Our testing environment consisted of two testbeds. The first testbed consisted of two server-class Athlon systems with a 266MHz FSB. The second testbed consisted of typical desktop/workstation Pentium-based systems.
Twin Server-Class Athlon systems from QLITech Linux Computers
Tyan S2466N Motherboard
AMD 1500MP
2x64-bit 66/33MHz jumper-able PCI slots
4x32-bit PCI slots
512MB DDR Ram
2.4.17 Kernel
RedHat 7.2
Twin Desktop-Class Dell Optiplex Pentium-Class systems
Pentium III 500 Mhz
128MB Ram
5x32-bit PCI slots
3x16-bit ISA slots
Our tests focused on aspects of
Throughput
Latency
Cost per Megabit
D-Link DGE-500T
.0002 seconds.
/((192.21+172.21) / 2) = $.25>
.0002 seconds.
.0002 seconds on both test platforms.
D-Link DGE-500T was the first of the gigabit cards tested. This card is based on SMC's dp83820 chipset and is designed for a 32bit bus. The chipset in this card turned out performance nearly identical to the two Ark cards and the GigaNIX cards tested in our test suite, since all utilize the dp83820 chipset from SMC. The Linux driver used was the ns83820 as included in the 2.4.17 kernel. Latency on both platforms was
Peak throughput while operated in a 32bit bus was 192.21 Mbps. This was achieved in the Dell systems. The Athlon systems only obtained a peak of 172.21 Mbps when these cards were inserted into the 32-bit bus. Both systems show a slight drop in throughput but eventually level out. Peak throughput while operated in a 64bit bus running at 33Mhz was 315.96 Mbps.
When the bus was jumpered to autoselect 66/33Mhz, the performance increase was negligible. Peak throughput was 316.40 Mbps. Comparing the plots of the 66Mhz and 33Mhz run reveals that they are essentially identical.
For complete testing results, click here.
Price: $45
The cost per Mbps is as follows:
32bit 33Mhz: $45
64bit 33Mhz: $45 / 315.96 = $.14
64bit 66Mhz: $45 / 316.40 = $.14
Ark Soho-GA2500T
The Ark Soho-GA2500T is also a 32-bit PCI card design. Like the D-Link DGE-500T and the Asante GigaNIX cards, this card is based on the SMC dp83820 chipset. With that in mind the performance was estimated to be close to the D-Link DGE500T. The driver used was the generic ns83820 included the 2.4.17 kernel. The latency for both test systems was
The peak throughput achieved while in a 32bit 33Mhz bus was in the Dell system: 192.62 Mbps. While the Athlon system in the same bus setup only reached 172.19 Mbps. As before, there is a performance drop at the 1Kb and 5-10Kb packet sizes.
Peak throughput while operated in a 64bit bus running at 33Mhz was 610.83 Mbps and 609.98 Mbps when running at 66Mhz respectively. As with the Soho-GA2000T, there is no noticeable difference between a 33Mhz and a 66Mhz bus.
For complete testing results, click here.
Price: $44
The cost per Mbps is as follows:
32bit 33Mhz: $44 / ((192.62+172.19) / 2) = $.24
64bit 33Mhz: $44 / 610.83 = $.07
64bit 66Mhz: $44 / 609.98 = $.07
Ark Soho-GA2000T
Our transition into cards designed for a 64-bit PCI bus began with the Ark Soho-GA2000T. Like it's 32-bit counterpart, this card was designed around the ns83820 chipset, which will allow us to examine the performance benefits, if any, in moving from a 32-bit As
Designed to run in a 64bit 66Mhz slot, this card is backwards compatible to 32bit and 33Mhz slots. This card is based off of SMC's dp83820 chipset so performance was expected to be similar to the DGE500T and the Soho-GA2500T. The driver used was the generic ns83820 included in the 2.4.17 kernel. Latency was
Peak throughput for a 32bit 33Mhz slot was 189.93 Mbps in the Dell system. The Athlons were only able to reach 172.26 Mbps.
Peak throughput for 64bit 33Mhz was 665.06 Mbps with an MTU of 6000. Peak throughput while running at 66Mhz was 640.60 Mbps. With the exception of the 6000MTU tests, there is no noticeable difference between bus speeds of 33 and 66Mhz.
For complete testing results, click here.
Price: $69
The cost per Mbps is as follows:
32bit 33Mhz: $69 / ((172.26+189.93)/2) = $.38
64bit 33Mhz: $69 / 665.06 = $.10
64bit 66Mhz: $69 / 640.60 = $.11
Asante GigaNIX
.0002 seconds on both systems.
.000048 on the Dells and .000025 seconds on the Athlons. Of all the 64bit cards tested, the SK9821 is the first to have a noticeable difference in performance between the two bus speeds.
.000123 seconds.
.000103 in the Dells and .000078 in the Athlons.
.000091 seconds. Due to time constraints, we have yet to test this card in the Dell testbed.
.4 Mbps for several TCP packet sizes. At an MTU of 6000 and at 9000 the same problem occurred as before in the 64-bit 33Mhz test.
The second 64bit card tested was Asante's Giganix. This card is designed for a 64bit bus but, is backwards compatible to 32bit and 33Mhz configurations. Giganix is based off of the dp83821 chipset. The driver supplied by Asante was unable to compile due a bug in the code. In order to get the card to work the generic ns83820 driver was used again. Performance was expected to be similar to the GA2000T. Latency was
Peak throughput for a 32bit 33Mhz configuration was 238.75 Mbps in the Dell systems, with a peak of 172.19 in the Athlons. When comparing to the GA2000T, the Athlon results stay about the same whereas the Dell systems increase by 50Mbps.
Peak throughput for 64bit 33Mhz 641.02 Mbps with an MTU of 6000. When running at 66Mhz, the peak is 651.51 Mbps with the MTU at 6000.
An interesting spike in throughput on the 64bit 66Mhz tests was when the MTU was set to 3000. Aside from the 40Mbps difference between the two bus speeds, the plots look very similar. The main difference is the spike at 8KB packets.
For complete testing results, click here.
The cost per Mbps is as follows:
32bit 33Mhz: $138 / ((238.75+172.19) / 2) = $.67
64bit 33Mhz: $138 / 641.02 = $.22
64bit 66Mhz: $138 / 651.51 = $.21
Syskonnect SK9821:
The first of the Syskonnect cards tested was the SK9821. This card is designed for a 64bit bus. The SK9821's are backwards compatible to 32bit and 33Mhz configurations. The driver used was sk98lin from the kernel source. Latency was
Of all cards tested, the Syskonnect SK9821 gave the most consistent throughput over all packet sizes, and was far-and-away the overall performance leader.
In the server-class testing environment, peak throughput in our 64 bit 33Mhz setup was 782.27Mbps with the MTU set to 9000. The peak for 66Mhz tops off at roughly 940Mbps with jumbo frame MTU sizes of 6000 and 9000.
Peak throughput on 32bit 33Mhz was 365.27 Mbps on the Dells. After the peak, is reached there is a noticeable drop in throughput as it levels off to the 330Mbps range.
For complete testing results, click here.
Price: $570
The cost per Mbps is as follows:
32bit 33Mhz: $570 / ((365.27+163.97) / 2) = $2.15
64bit 33Mhz: $570 / 782.27 = $.73
64bit 66Mhz: $570 / 938.97 = $.61
Syskonnect SK9D21:
The second card tested from Syskonnect was the SK9D21. The SK9D21 is aimed at the desktop/workstation market. While support for this card under Windows environments appears to be solid, there were too many technical issues. The testing environment's mix of kernel, motherboard, Athlon chipset, and Syskonnect drivers made for too many components to successfully debug the problems with this card thoroughly. This card is designed for a 64bit bus the card is backwards compatible with 32bit and 33Mhz configurations. While an exhaustive analysis of the cards was unavailable, it should be noted that the latency was successfully determined at
Our difficulties with this card were limited to the 64-bit bus. Our tests were successful in analyzing the performance in both the QLITech Linux Computers Athlon-based systems and the Pentium-based systems in 32-bit busses.
When drivers issues for this card are resolved, performance evaluations in this section will be amended.
Peak throughput in the Dell system was 377.53 Mbps. As with the SK9821, there is a drop off after the peak is reached.
For complete testing results, click here.
Price: $228
The cost per Mbps is as follows:
32bit 33Mhz: $228 / 377.53 = $.60
3Com 3c996BT:
The next card in the test suite was the 3Com's 3c996BT. This card is designed as a 64bit 133Mhz card, but is backwards compatible to 32 bit, 33 and 66Mhz configurations. The driver used was the bcm5700, version 2.0.28, as supplied by 3Com. Latency was
The peak throughput achieved in this card while in a 32bit 33Mhz slot was 436.23 Mbps in the Dell systems. In the Athlon system, the same bus configuration only reached 184.02 Mbps.
Peak throughput while running in a 64bit 33Mhz slot was 884.09 Mbps this was with an MTU of 4000. While running at 66Mhz, the peak was only 546.16 Mbps with an MTU of 6000. These plots are all relatively smooth when compared to the other plots for this card.
Performance in a 66Mhz slot is actually lower for all MTU sizes as compared to a 33Mhz slot.
For complete testing results, click here.
Price: $138
The cost per Mbps is as follows:
32bit 33Mhz: $138 / ((436.23+184.02) / 2) = $.44
64bit 33Mhz: $138 / (884.09) = $.16
64bit 66Mhz: $138 / (546.16) = $.25
Intel Pro 1000/XT:
The final 64bit card tested was Intel's E1000 XT. As with the 3c996BT this card is designed for future PCI-X bus speeds running at 133Mhz. It is compatible with a variety of configurations running at 33 and 66Mhz as well as 32bit. The card uses Intel's e1000 module, version 4.1.7. Latency in the Athlon systems was
Peak throughput achieved was 743.14 Mbps while running in a 64-bit 66Mhz slot with the MTU set to 9000. Performance in a 32-bit configuration turned out the lowest throughput for all cards tested coupled with the most erratic throughput. During the throughput tests, the card would drop 100% of packets for extended lengths of time. Initial testing in the 64-bit setup showed performance similar to the Giganix card with regards to a 64-bit bus. Once the MTU was set to 9000 performance became very erratic, stagnated several times, then stabilized once the packet size reached an upper threshold peak. Note that the drop in performance was not associated with the (expected) phenomena of packet reassembly when the TCP packet size exceeds the MTU.
As testing continued the the 66Mhz phase things only got worse. Once the MTU exceeded 3000, performance was no longer predictable. During the 4000 MTU tests, the throughput plummeted to around
For visual clarity of this phenomena, see the ''Complete Test Results'' link for the Intel Pro 1000/XT below.
For complete testing results, click here.
Price: $169
The cost per Mbps is as follows:
32bit 33Mhz: $169 / 142.02 = $1.18
64bit 33Mhz: $169 / 624.41 = $.27
64bit 66Mhz: $169 / 743.14 = $.22
Just fyi, Macintosh 1000BaseT ethernet controllers go directly to the memory controller, bypassing PCI altogether..
Care about electronic freedom? Consider donating to the EFF!
Stay away from cards that don't have PXE and cards in which the driver won't compile into the kernel (as opposed to a module) if you plan to do easy installations or mount root off the network. In other words, stay away from Netgear and some 3Com cards (I haven't tested others), and play it safe with Intel.
Gigabit, not GigaByte. Gigabit = 1000000000 bits. GigaByte = 8000000000 bits. A 1 GBps connection is 8 times faster than a 1 Gbps connection.
Well, check out the docs first off. It's hard to get much out of GBit, since most of the utilities don't call the socket open with properly sized buffers/window/whatever.
/* 8192 */
I set up optical gigabit for some NAS type things at work, and out of the box, GBit performed maybe 30% better than 100 Mbit. We are talking about 110Mbit peaks, compared to 80Mbit peaks with 100Mbit switched.
Setting the MTU to 6144 (max that I could set it to with the ns83820.o) I started to get peaks around 300Mbit/sec.
I tried recompiling the module for higher limits, since in the source it has:
#define RX_BUF_SIZE 6144
But if I put in 8192, or 9000 like I wanted it to be, it would crask or lock up.
Anyway, it's not trivial to get good performance out of GBit, and definitely don't expect anywhere near 10X gain.
I've had enough abrasive sigs. Kittens are cute and fuzzy.
Comparisons and Observations
In this section, we compare performance differences between cards in like environments , provide some general performance observations, and examine the cost per megabit as determined by the operating environment.
Head-to-head throughput results
While the results obtained in this study clearly show that peak performance is not a complete indicator of peak performance, in this section we examine the peak performance results amongst all cards under common environments.
32-bit, 33MHz PCI Bus, 1500 MTU
64-bit, 33MHz PCI Bus, 1500 MTU
64-bit, 66MHz PCI Bus, 1500 MTU
64-bit, 33MHz PCI Bus, 3000 MTU
64-bit, 66MHz PCI bus, 3000 MTU
64-bit, 33MHz PCI bus, 4000 MTU
64-bit, 66MHz PCI bus, 4000 MTU
64-bit, 33MHz PCI bus, 6000 MTU
64-bit, 66MHz PCI bus, 6000 MTU
64-bit, 33MHz PCI bus, 9000 MTU (Note: Drivers for the dp83820 chipset were limited to around 6000 MTU)
64-bit, 66MHz PCI bus, 9000 MTU (Note: Drivers for the dp83820 chipset were limited to around 6000 MTU)
General Observations
Of the eight cards tested, the clear performance champion was the SK9821 with regard to throughput and consistency. The 3Com 3c996BT has a modest price tag and respectable performance for the entry-level server configuration. If price per megabit is the main concern, the Ark Soho-GA-2500T has the lowest cost per Mbps, making it a viable solution for entry-level systems requiring higher throughput than fast ethernet.
The D-Link DGE500T and the Soho-GA2500T show nearly identical peaks, which is to be expected since the drivers and the chipsets were the same.
The 3Com 3C996BT has results when compared to the 64-bit 33MHz results were surprising inasmuch as these cards showed better performance at 33MHz bus than at the higher 66MHz bus.
Of all of the cards tested, the Intel E1000 TX proved to be comparable to the comparable to the Asante GigaNIX card in peak performance, but the erratic overall performance proved too much to overcome.
In referring to the Complete Test Results sections for the 3C996BT and the SK9821 cards, one sees a very consistent and smooth transition to the peak throughput of the cards over the complete range of packet sizes.
Some general comparisons that can be derived from the above results include the notion of ''cost per peak megabit. Depending upon the environment that the network device is to be installed, the cost per peak megabit varies greatly. For example, if one would wish to upgrade their P-III-based desktop system with a 32-bit, 33MHz PCI, the GA25000T is the clear cost-effective solution, but would not be able to provide throughput at the level of the 3Com 3C996BT.
In an HPC environment, where sustained throughput is critical and the switch is capable of Jumbo frames, the SK9821 would be the best performer. In light of gigabit switching hardware that lacks Jumbo Frame support, a comparison of the 1500MTU results shows the SK9821 is still a viable choice, as is the 3Com 3C996BT which provides a more cost-effective solution.
Paul Gray
323 Wright Hall
University of Northern Iowa
Well... copper is cheaper than fiber for the moment. I'd hate to think what my 50 meter run from my router to the second floor of my townhouse would cost if it was fiber.
I use optical runs for my audo as well, but those are all under a meter, for the most part, and around $30 or so a piece. Not too much money for the purpose, but I dont' think I'd enjoy paying for a 50 meter run. Never mind the cost of devices with optical interfaces.
That said, I guess the only reason I'd consider GB copper is that it's no more expensive than 100 base-T...
"Oh my God. This is terrible. This is the end of my Presidency. I'm fucked."; ~ Donald J. Trump
apparently Pricewatch.com has D-Link 8-port 10/100/1000baseT auto-detect switches listed for under $150!
;).
These are for 8x100-base-T with a gigabit uplink. I researched this a while ago, when speccing out my dream network
The cheapest full-gigabit switch D-link sells is about $1500.
The cards are well priced for home use, and CAT5E cabling is cheap too. The problem with gigabit ethernet is not the cards, it is the lack of switches or even plain hubs at an affordable price point. There are lots of switches out there with a single gigabit port, but even those are a couple of hundred dollars. If you want multiple gigabit ports, you are looking at more than $600 for the bottom rung products.
When information is power, privacy is freedom.
Someone needs to learn the difference between a gigabit and a gigabyte....
Doing Math we can calculate that a full gigabit of transfer is 125 Megabytes a second. I think this is possible with high end hard drive technologies like SCSI RAID and speeds like this will probably show up in the desktop in a few years.
And, of course, not all data has to be written on Hard Drives. You could have a router or switch that will pass along a gigabit of packets a seconds but it certainly doesn't write them to the hard drive. You could for example but in a Gigabit Ethernet connection between two nearby buildings.
Tim
Omnia vestra castrorum habetur nobis.
I got about 32 MByte/s one-way with `ttcp` [UDP] between a 1.2GHz K7 and 2*500 Celeron (BP-6) through a plain crossover cable.
Not bad, but only 25% of wirespeed (125 MByte/sec). I figured the main limit was the PCI bus, which would only burst at 133 MByte/s, and I strongly suspected that the bursts were too short to achieve anything like this speed. I have yet to play with the PCI latency timer.
One thing for sure -- it isn't the CPU speed or Linux network stacks. The K7 will run both ends of ttcp through the localhost loopback at 570 MByte/s, and the BP6 around 200 MB/s.
>> Gigabit ethernet over copper has the advantage of running over your existing cabling (i.e. cat-5 is fine). This avoids having to muck about with fiber, as fiber is a PITA to maintain yourself (getting optically perfect connections for the fiber jacks is picky).
Actually, the siecor unicam series work really, really well. They use a index of refraction matching gel inside the factory polished terminators. All you have to do is cut'n'crimp. They work great. I haven't ever had to do any splicing though - but, given how well the siecor stuff works, I can't see it being a remarkable problem.
fnord.
There was another review of GigE performance in the IEEE Network Magazine last year.
D-Link's site is nearly impossible to navigate (maybe it requires JavaScript, which I've shut off), but the Pricewatch description of the DES-1009G indicates that Gigabit Ethernet is only available on one port as an uplink connection; the rest of the switch is your run-of-the-mill 10/100 job. The DGS-1008T is D-Link's 8-port unmanaged 10/100/1000 switch; the cheapest entry on Pricewatch for that is $595.
BTW, I have the entire site downloaded. Maybe I'm insane to even think about mirroring a /.'ed article on my home cable-modem link, but here it is. I've converted all the charts to PNG so they'll load slightly faster, and I got rid of most of the godawful "super-31337" yellow-on-black text to improve readability. You can also choose this link to download the entire page (images and all) in one shot.
20 January 2017: the End of an Error.
Technically we are both "right". However, when you use the term "bits" you mean binary digits. emphasize binary. We are not using base 10, but base 2.
The Free On-line Dictionary of Computing (13 Mar 01)
prefix
1. The standard metric prefixes used in the SI
(Syst`eme International) conventions for scientific
measurement. With units of time or things that come in powers
of 10, such as money, they retain their usual meanings of
multiplication by powers of 1000 = 10^3. When used with bytes
or other things that naturally come in powers of 2, they
usually denote multiplication by powers of 1024 = 2^(10).
Here are the SI magnifying prefixes, along with the
corresponding binary interpretations in common use:
prefix abr decimal binary
yocto- 1000^-8
zepto- 1000^-7
atto- 1000^-6
femto- f 1000^-5
pico- p 1000^-4
nano- n 1000^-3
micro- * 1000^-2 * Abbreviation: Greek mu
milli- m 1000^-1
kilo- k 1000^1 1024^1 = 2^10 = 1,024
mega- M 1000^2 1024^2 = 2^20 = 1,048,576
giga- G 1000^3 1024^3 = 2^30 = 1,073,741,824
tera- T 1000^4 1024^4 = 2^40 = 1,099,511,627,776
peta- 1000^5 1024^5 = 2^50 = 1,125,899,906,842,624
exa- 1000^6 1024^6 = 2^60 = 1,152,921,504,606,846,976
zetta- 1000^7 1024^7 = 2^70 = 1,180,591,620,717,411,303,424
yotta- 1000^8 1024^8 = 2^80 = 1,208,925,819,614,629,174,706,176
------
BINARY BINARY BINARY! WE USE BINARY! Take a look to the right under "mega". mega- M 1000^2 1024^2 = 2^20 = 1,048,576. Therefor, 2^30 / 2^20 = 2^10 = 1024 megabits in 1 gigabit.
Now, what part of that dont you understand?
Buying a Dell computer is equivalent to dropping the soap in a prison shower.
The fastest pci gets is 66mhz 64bit. Thats 64 bits per clock cycle, 66M clocks per second....4.224 Gigabits. I'd say thats a little higher than 1 Gb.