Originally Posted by qasdfdsaq
And the bigger picture, analyzed from a point of view other than the haze of VM's "uber duper fibre optic super" marketing, important points bolded for emphasis.
First of all, the report is based on Samknows' broadband speed testing hardware and technology. As such, most speed-related tests only measure download speed from certain fixed targets and not all-round performance. It is well know that VM run certain peering links at close to or at maximum capacity causing performance problems to/from certain destinations while others are fine (Youtube is a commonly reported issue). Hence, just because you get a 50mb download test doesn't mean you'll get 50mb from the site you want to. The tests purposefully exclude heavy users and congested areas, and therefore somewhat bias VM by excluding anyone whose connection speeds would be regularly reduced by STM - a technique other networks do not employ.
OK, let's begin.
95% of consumers are still on 24mb or lower connections - i.e. less than 5% are on FTTC, FTTP, VM 30mb, 50mb and 100mb combined. (1.2)
The average speed of VM's 50mb product has actually fallen compared to the previous six months - though not in a statistically significant manner, and are still much higher than a year or two ago. Yet it is now exactly equal to BT FTTC as a proportion of provisioned speed. (1.3) The argument that you get more of the speed you pay for on cable (vs. xDSL) is no longer completely true in reference to VDSL.
Average speeds have decreased on most VM packages from May to Nov 2011, while on xDSL products it has improved in all bands.
BT FTTC performance ("the competition") has improved drastically over the last 6-12 months, particularly at peak times. (1.4)
The average BT "fibre optic" connection is over 2.5x as fast than the average VM "fibre optic" connection. (1.5)
Over 50% of customers receive at least 95% of their advertised speed on both cable and FTTC. The 25th percentile on FTTC is only 2% lower as a proportion of provisioned speed vs 50mbit VM cable. (1.6)
VM cable still loses almost twice as much of their peak speed at busy times compared to BT FTTC. FTTC customers lose barely 4% at peak time, VM 10mb customers lose 10%, 30mb lose 4.5% and 50mb customers lose 9% (1.7). In spite of having the highest headline speeds, VM connections also still suffer the highest congestion.
Again, BT's 20 and 40mb packages improved significantly over the last 6-12 months, none of VM's 10, 30 or 50mb packages showed any such improvement. In other words, download speeds on BT are catching up fast (and considerably beat all but VM's 50mb and 100mb services). (2.6 & 2.9)
In terms of upload speed, 40/10 FTTC's actual performance is considerably higher than any of VM's packages surveyed. Actual performance of BT Infinity is 30% slower download but over 80% faster upload than the fastest VM connection surveyed. (2.15)
VM cable also still has the highest difference between peak speed and average speed than FTTC - again and indicator of susceptibility to congestion. (4-6) Contention and congestion on FTTC is still far lower than on VM.
Both average and peak time website loading speeds are faster on BT FTTC than any of VM's packages - including 50mb. Both VM 30mb and 50mb get slower at peak times, while BT FTTC actually gets faster (34). Even with a slower download speed, actually opening web pages is around 20% faster on BT 40mb than VM 50mb.
Average latency (ping) is lower on BT's 40mb than on any of VM's packages surveyed (including 10, 20 and 50mb). (36 & 40). Along with much faster upload speeds, this makes FTTC much better for gaming than VM, contrary to VM's claims. Also, latency on VM is roughly the same regardless of whether you're on 10, 30, or 50mb. Upgrading to a "faster" cable connection won't improve your gaming latency at all.
The increase in ping at peak times is much higher on VM than on FTTC. In other words, another sign of congestion on VM's network (36 & 40) and another red mark on VM's gaming performance.
Packet loss on VM still remains higher than the competition, though the differences are much smaller now on higher tiers. Despite this, packet loss on VM is still up to 2-3x higher than some of the competition (e.g. BT, Sky) across all tiers, particularly worse at peak times. (42 & 46)
VM's use of anycast DNS used to give them faster DNS lookup times than everyone else, but BT Infinity is now even faster (52). DNS lookup times on BT FTTC are 30% faster than on VM. Along with lower latency, this explains why websites load faster on BT than VM even when you pay for "more speed" from VM.
Jitter, another measure of gaming and VOIP performance related to latency, is still 2-4x worse on all VM services than even 8mbps DSL. Upstream jitter on VM is ten to twenty times worse than on BT FTTC. (64) Downstream jitter on VM is significantly worse than the competition on the 10mb package, but not significantly different on 30 or 50mb. (70)
Again though, rather oddly BT FTTC's performance (in terms of jitter) actually improves at peak times, while VM's gets considerably worse (64 & 70).
That brings us to the end of my analysis of the results. As a disclaimer, yes these are a very selective and polarized view, but someone has to take the opposite of "VM has got the best".
As an overall analysis of all the metrics above, VM only win outright in terms of peak download speeds. In all other measures they are now worse than BT Infinity, partly because BT Infinity has improved drastically while VM's service(s) have not. As a consequence, anything aside from downloading large files is unlikely to perform better on VM. Even actual web browsing is faster on BT 40mbit than VM 50mbit.
|