V.90 Interoperability Status

V.90 Interoperability Status (Updated 29-Aug-02)

29 Aug 02 - While (usually minor) bugs still crop up in client to server modem interoperability, for the most part, V.90 has matured and works - on 56k-compatible lines. The biggest problem is the myriad of client and server modem drivers and firmware. Most modems are sold with old drivers - and many problems can be solved by getting an updated driver. End-users have little control over the server-side, and ISPs must contend with various bug fixes, and new ones that crop up in the various releases available for their equipment. Now, the problem becomes V.92: these new modems, and the associated drivers and firmware have their own interoperability problems; and, in some cases, ISPs find that new V.92-enabled upgrades introduce some new V.90 interoperability problems.

Also note - as you read all the info on this page, understand the time frame: as you go down, the time-frame of the text gets older and older.

8-Apr, '00 - Nortel Networks CVX1800 access switch is being deployed by some ISPs and CLECs. This presents another possible V.90-server modem to deal with. My experience has been horrible: I cannot get a good 56k rate, and neither of my modems (USR-Sportster and LT Win Modem) can maintain a connection to the CVX1800 deployed by CLEC/ISP GST in Honolulu for more than a few minutes. It's possible that the problems I'm seeing are due to provisioning problems (line coding) - or it could be the firmware in the new CVX.

In spite of having a good 56k-compatible line (straight copper ~8000 feet to a remote office trunked to switching CO with no RBS), I am experiencing severe connectivity problems: my present ISP uses UUNet-provided local access and I experience intermittent loss of DNS/PPP without disconnection, disconnections, retrains, and a few 56k-connect failures. Providers using UUNet POPs include GTE, MSN, Dot.now, SurfBest. UUNet operates over 2000 POPs with Ascend TNT equipment - and, some of them work well some of the time. Prodigy, as well as 1stUp and others use SplitRock's Bay Networks-powered access POP. It gives me most of the problems I get with UUNet and a lower connect rate and throughput and pings can be abysmal. A growing number of local ISPs use the CVX as described above - unusable. I recently needed to download a 110mb text file from a ftp server that doesn't support broken downloads. I tried over 20 times via UUNet, SplitRock, and the CVX, and once got 90mb of it. I opened a trial account with AT&T (using a local number with 3Com Total Control equipment) and got the file on the second try with excellent throughput. Problem is: sooner or later, the AT&T POP will probably encounter [temporary?] performance impairments --- there's just too much that can go wrong.....

1-Feb, '00 - Interoperability problems continue - and perhaps are increasing. ISPs operating 3Com server equipment - which on the last update I called the most connectable - are complaining that owners of inexpensive software modems (PCTel, Soft56, etc.) are able to get only a 26.4k connect to their equipment, but can get 40k or better to Ascend server equipment. 3Com appears to be out-of-touch with its customer base - both end users and ISPs. Lucent which now owns both Ascend and Livingston has indicated it will not continue to produce new models in the Livingston Portmaster line - favoring the AscendMax line for ISPs. While the Ascend equipment may connect better to some of the cheapest (and worst) 56k client modems, in nearly all cases, Ascend will do very poorly for the upstream V.34 rate on a V.90 connect - 21.6k or 24k. If the 3Com gear hooks up at a 56k rate, it is still likely to yield a better upstream rate with any brand of client modem. If the client phone line link to the ISP has no RBS links, and if the ISP is connected to the telephone network via ISDN (as opposed to channelized T-1), and if there are no line coding mis-matches, and if there are no extra A/D conversions, a user is likely to achieve a 40k or greater connect with almost any V.90 client modem calling almost any V.90 server modem. But, add any of these factors - many of which are common - and 56k interoperability problems are likely to occur. ISPs that opt to connect to the phone network with [less expensive] T1 facilities or CLECs (competitive local exchange carriers) are in many cases asking for trouble - poorer connectivity than if CLECs and T1 are avoided. On the client side, the Lucent Win Modem may now be the most connectable modem as the latest LT firmware is able to achieve 56k connects when there are 'non-standard' RBS links. Motorola and PCTel modems (which use the PC's CPU to do all the processing) can be had new for $8 to $15 or less - and they aren't even worth every penny of what you'll pay (for most people - they are junk); Rockwell/Conexant Soft56 (also a HSP modem) may have a slightly better chance of working - but not much. The 3Com Internet Gaming Modem is a modem made by the marketing department - not engineering - and doesn't live up to the promise of faster gaming. 56k 2 years after V.90 modems hit the market is still elusive for 30-40% of modem owners - and this site remains one of the best sources of reliable information about this still-not-ready-for-prime-time technology.

Aug 28, '99 - Significant improvements have been made to firmware for many client modems as well as for 56k server modems that improve V.90 interoperability. While the end-user can upgrade (and in most cases downgrade) their modem firmware, the ISP they use has to decide which of the many firmware versions - all with their own unique issues - to install. ISP firmware changes will affect 56k connectivity for a sub-set of users. In addition, changes made by your telephone company in their facilities and network can cause 56k to start - or stop - working.

While 3Com/USR has made some of the best V.90 server and client modems, they are losing money in the analog modem business, and have diverted significant resources to future broadband products. 3Com modems are becoming 'less-connectable'. Lucent continues to improve firmware for their products. Rockwell/Conexant appears to be making some effort, but still comes in at #3.

This week, a "study" was released by Inverse that says V.90 is now performing as promised. (See Update- 27-Aug-99) This study suggests that V.90 works over 93% of the time. But, I estimate that 30-40% of 56k modem owners currently don't get 56k connects at all. Of those that do get 56k connects, a significant number have unreliable connections with high error rates, retrains, and disconnections. In many cases, the throughput achieved on a 56k connection equates to a connect rate 10% - 50% lower than the reported connect rate. (The Inverse study found 56k throughput to average 38k, while connect rates averaged 46k.)

Many modem owners can improve their connections by upgrading firmware, and/or using speed limit and protocol selection commands. See the Troubleshooting Section.

May 1, '99 - There is little change since last (Mar'99) update. Interoperability problems continue.... meaning, some users of some modems calling some ISP server gear using some telco lines are unable to achieve a 56k connection - but, can achieve it with another brand of modem, or calling different ISP server modems. On the other hand, there are a certain amount of lines and digital telco facilities preventing all 56k connects. See What's a 56k-compatible line? It's hard to determine exactly what works and what doesn't (there are ever-increasing possible combinations of firmware...), but here's my sense of the current state of affairs:

3Com continues to have the "most connectable" client and server modems - but they're far from perfect. Some 3Com server releases are 56k-incompatible with some 3Com client modems; Lucent-based modems with the latest firmware do well when calling 3Com or Lucent/Livingston server modems, but may have problems calling Ascend or Cisco-based server modems. Rockwell-based modems do best calling Ascend or Cisco servers, and may have trouble, or lower connects, calling all others. With any client modem, you're likely to see a reduced upstream rate (ie, 26.4 or lower) calling anything other than a 3Com server - you're likely to get a 26.4 or higher rate calling a 3Com server. Ascend has recently released new modem firmware that may improve Lucent->Ascend interoperability. All vendors of server modems have been releasing new code - a big factor in whether your modem will get a [decent] 56k connect. Prudent ISPs are often reluctant to install new code in their modems - all releases have their own issues. Rockwell/Conexant HCF modems - now included in many new computers, are problematic and owners with connectivity problems should check for firmware/driver updates. PCTel, ESS and Cirrus/Ambient modems also suffer from bad firmware/drivers.

11/18/98 - Since the last update to this page, new server-based firmware has been made available for ISP equipment including 3Com HiPer, Cisco AS5200/5300, and Bay Networks. A variety of interoperability problems still exist - as well as new issues resulting from the newest server firmware releases. Most interoperability problems depend to some degree on the telco facilities involved on the call (which can vary from call-to-call), as well as the client and server modems. This makes it extremely difficult to identify and correct the problems. In addition to 56k connectivity issues, there are continuing problems making v.34 connections to some 56k server modems.

9/18/98 - A number of ISPs using the latest 3Com server 'modems' - the HiPerARC/DSP - are reporting that users of Lucent LT Winmodems can no longer connect (the modems do an 'endless handshake'), and that they didn't have the problem until they upgraded and replaced their 3Com Quad modems. Another problem of premature disconnects is also being reported. It is clear this is an interoperability problem - it is unclear as to whose problem (Lucent or 3Com) it is. 3Com is promising an ER (engineering release) of code for its server-side products "in the near future" to address as many of the Rockwell and Lucent interoperability issues that it is able to with fixes to 3Com server code. (11/98) This problem is fixed with Lucent firmware 5.22 & later.

9/6/98 - I check and present results of V.90 test calls from 2 different lines (in different telco offices) with 4 different modems, to 5 different ISP server #s using 4 different makes of server hardware. See the results here.

3Com has V.90 client modems and V.90 server firmware released. (Most 3Com/USR x2-modems have a V.90 firmware upgrade available.)

Rockwell-based V.90 "dual-mode" modems are on the market. (Most vendors have released "upgrades" for Rockwell-based Flex modems.)

Ascend V.90 server firmware has been released.

Lucent/Livingston has V.90 server firmware has been released.

Lucent V.90 client modem V.90 firmware upgrades are available. (See the LT Winmodem page.)

STATUS:

3Com client to 3Com server: V.90 will work in many situations. It will allow 56k-connects for some people who could never get them (on local calls) with x2. There are situations where V.90 fails; as well as situations where the CONNECT rate as well as throughput are significantly less than the user can achieve with x2.

Any client to Bay Networks (x2 or V90) server: Prodigy is the only major provider using Bay equipment. Many problems have been reported. I suspect some of the problems are related to telco issues - Prodigy/SplitRock primarily use CLECs and super-POPs for access, which can cause additional complications. Prodigy/SplitRock are rolling out V90, and it appears some CLEC and telco routing issues will persist. Bay equipment will connect to V90 clients, but the upstream rate may be much lower than with calls to other servers. It appears that Bay is the only server vendor other than 3Com that supports SREJ (selective reject)*. As of 11/98, the Prodigy/Bay Networks V90 rollout is proceeding very slowly as a variety of connectivity/interoperability problems surface. See the Prodigy Network Status Page.

Rockwell client to 3Com server: V.90 will work in many situations. My limited testing with a Zoom 2949 dual-mode external modem indicates that the CONNECT rate and throughput can be about 20% less than a similar connection with a 3Com client modem. Latest Rockwell client firmware releases appear to be less capable of 56k connects in some situations! (Like mine where the Zoom is useless.)

Rockwell client to Lucent/Livingston server: V.90 will not work. The modems will negotiate a Flex connection (if the modem is dual-mode) [9/6/98 - this may or may not still be the case.]

3Com client to Lucent/Livingston (beta) server: V.90 will work in many situations. In my testing, the CONNECT rate will be close to that achieved calling a 3Com V.90 server, however the transmit side of the connection may be 10-20% lower than the 3Com achieves on a call to a 3Com server. The throughput is much lower than the CONNECT rate indicates:

It appears the Livingston server modems have a problem with speedshifts, and null speedshifts, as well as with error recovery: I did a FTP download of a compressed file local to the PM3 and got a throughput of around 3.6KB on a 46.6kbps connection. I noticed that whenever there was a speedshift or null speedshift, data flow from the server would cease for as much as several seconds; it also appears that re-transmission as the result of blers can cause a much more pronounced drop in throughput compared to a 3Com-3Com connection. (See 4/25 update for more info on null speedswitch.) Another factor may be that Livingston, like Ascend, does not support SREJ. (Selective reject.)*

Lucent clients: V.90 working calling 3Com servers. May or may not do V.90 to Livingston and Ascend servers. With latest Livingston servers, V90 appears to work in some cases; however, KFlex must be disabled when calling Ascend servers, otherwise only a KFlex connect is negotiated.

Lucent/Livingston servers:
Lucent/Livingston servers: Livingston V90 was officially released for the Portmaster equipment just as the ITU gave its official blessing to V90 (9/98). The V90 modem code in the Lucent release was unchanged from the last (21st) beta version despite a range of outstanding connectivity/interoperability issues. Lucent continues to work on new firmware for its server modems - which (11/18/98) still hasn't been released. ISPs using Portmasters continue to complain about a number of problems including v.34 and V.90 connectivity problems. Issues include premature disconnects, speed issues, ability to establish connection issues, as well as 'spontaneous' reboot of the Portmaster server.

Cisco servers: There are 2 server modems available for Cisco access gear - Microcom (Rockwell-based) and Mica (Cisco's own). Mica has only recently (~11/98) had V90-capable firmware; the Microcom modems have had V90 for some time. While Cisco recently claimed they have become the #1 supplier of dial-up access servers, I have been unable to find any major US-based ISP that uses Cisco 56k access gear. It appears that there are some universities and perhaps non-US ISPs using Cisco access gear; and, it also appears that they are experiencing a range of connectivity/interoperability problems.

* SREJ - Selective Reject - Can improve throughput of connections with data errors as it reduces the amount of data that needs to be re-sent to your modem. Supported in all 3Com V.90 client modems; not supported in any other brands that I'm aware of. Only 3Com and Bay Networks servers support SREJ. Both the client and server must support SREJ. Update - Feb '00 - Lucent PM3 and PM4 now support SREJ with ComOS 3.9b27.