Excellent job. Im in the process of building/upgrading a machine. I picked up a 980ti on ebay for $79 hopefully I can get ultra 4k 60fps. Whatever I end up getting your tool will definitely come in handy. Thank you for your hard work and dedication. Donation is on its way! 👏🔥💯
Thank you! That's a good deal for a 980ti. Pair with a 12th gen Core i3 12100 or higher, or a CPU with equivalent high single core performance, to not be CPU bound in large groups. At 4K, leafy areas in Makuri Island might dip below 60FPS, but 980ti will be fine everywhere else.
Thank you for putting the effort and enthusiasm into both the tool and this video. Much appreciated 👍 Oh yes, the other runner will be along shortly I'm sure 😂
Excellent video and the tool itself is such a powerfull life saver. Great details, interesting stats! Thank you Mike for taking the time to share this stuff.
Fascinating video. Only came across Zwiftalizer recently when I was trying to think of how I could build/buy a PC affordably that would get me beyond Basic at average 3 fps on a good day! From your benchmarks had already deduced that an i3-121000 was probably the best buy, but what gpu? A top rate one was never going to be affordable, but after watching your video, I now don't think that I'll have to worry about this too much. Many thanks for this brilliant and instructive video.
Your tool is so useful ! It's like GODMODE 🙂 It helped a lot while looking for explanations on stuttering game, and options to make it better. I LOVE your approach on building a PC with used components for cheap ! Being in Europe I feel tricked by the 300ms ping... Many thanks for your time building the tool and the effort doing that video. They & you are very useful.
Great video, super interesting- thank you! One point to note is that my iPad Pro M1 runs basic profile - I think the Zwift look up table was only updated for the MacBooks for some reason!
Fascinating video; one thing that perplexes me is the affect of network latency on racing. Doesn't it mean people closer to the AWS server (people in the USA) are automatically at an advantage over the rest of the world and therefore gain an advantage in time critical aspects of the game, especially drafting. From the video there are obviously other issues that are in the mix, but how can racing and drafting be 'real' on Zwift?
No. There isn't any advantage. The server tracks each player's lag behind the central source of truth for what is "now," and modifies their time accordingly. Therefore, Zwift's in-game results are the only ones that are accurate. Other results systems that employ watchers located elsewhere will only be accurate in terms of their concept of "now" and if they account for each person's unique lag time. Pack dynamics, including position, rolling resistance, and draft, are computed client-side and sent, I believe, five times per second to the server. (What I said in the video about updating every 300 ms in not accurate - it updates every 200 ms regardless of what your lag time is). The fact that each player sees something different on their screen to where each rider is actually "now" according to the server is, in my opinion, a bigger problem. Being extremely close to the servers could actually work against you because you might believe that you were the first person across the line. In reality, though, someone who you thought was a few meters behind you might place first because the server subtracts their lag time from their finish time that you believed you were watching in real-time. In essence, it's close to real-time but never going to look accurate to everyone because of the physical distances the data has to travel are all different.
@@zwiftalizer Thanks for your reply. I think your last sentence is telling. ...if I follow you correctly racing on Zwift is a bit meh. I'm not saying it's not fun though..
@@timcorso6337 You summarised my feelings on racing accurately, and I would love to be proved wrong. It should be possible to achieve real-time gameplay with clever client-side position prediction and server time reconciliation. Maybe it's done, I just don't believe it's there yet. There's also a bunch of other factors that need to be solved for me to enjoy racing again, like sandbagging, weight doping, trainer accuracy and powermeter calibration fiddling.
@@zwiftalizer Hi, I am trying to decide whether to purchase a Zwift Ride to be used in a remote location using a mobile hotspot. I am concerned about latency with the Zwift spec being 50ms or less (my hotspot is about 150ms-175ms). In you help pages under "latency" you provide expected values of latency for places like London at ca 150ms, which I interpret as usable for Zwift but perhaps not "perfect", but contrary to Zwift-provided specs. You mention a 200ms update spec above so it would seem that a latency of 200ms or less would be compatible with a good experience on Zwift. I don't know that I would necessarily participate in Zwift racing but would like to ride through the winter season on the Zwift Ride just to keep a base level of cycling fitness/neuromuscular stimulus. I don't want to buy the Zwift Ride only to find that my 150ms latency is too jittery to have a reasonable experience on the platform. Do you have any guidance? Much appreciated and thanks for all of the work that you are doing!
Really great data set - really impressed with what you have created, I havent used it for a while so I will take another look after my next session! One question - you normalized the trainer type vs dropouts for numer of session, but have you normalized for session time? i.e. dropouts per 60 min of riding? Or was the distribution of different session times across different trainers close enough such that normalization was not required? Hope that was clear!
Thanks man. That's clear. Great suggestions. I'll take another look at this report. I will factor in session time, and take something like 5th to 95th percentile dropouts to exclude outliers.
I have an old Intel i5-4590 with a Nvidia 750ti from 2015, which works ok with Zwift. But now I'm looking at upgrading my computer. The CPU will probably again be in the i5 high end. As a not gamer person I don't know/understand GPUs. So, what I would like to get from your tool -- and tried to, but not really succeeded getting the few times I looked at it in the last years -- is to get some sense of which GPU I should buy that will give me Ultra in 4K in all conditions while not be overkill (as I don't play any other game), yet still have some room for future proofness.
Intel Core i3 12100F and NVIDIA 1660 Super GPU, 16 GB RAM, NVME SSD Gen3 is the current sweet spot for Zwift in Ultra 4K in (nearly) all conditions - including large groups, in all Worlds, almost everywhere. I say 'almost' because there are some graphically complex areas in Makuri Island that need a stronger GPU, but since those spots are few and far between it does not make sense costwize to go higher than a 1660 Super. Single core performance is important in large groups. More cores are not used effectively, so there is no benefit in going up to an i5 12th generation, but doesn't hurt and will get you some room for future proofness especially if you use the PC for other things.
@@zwiftalizer Hahaha, thank you for the suggestion! I did not mean to explicitly ask you for a recommendation here, I really meant that I would like if by looking somewhere in your tool I could easily figure this out by myself. But thanks again for the answer.
@@jp_baril you make a very good point - it is not easy to figure that out from the tool. The benchmarks for Makuri (being the most complex world) is where I look, but without costs, it is not easy to know which entires are good buys and which are overkill. I only know the i3 12100F/1660 Super combo is a good balance from talking to people. It is around entry 32 in this report zwiftalizer.com/benchmarks/9/ultra/2160/100 If costs, or popularity as a proxy for cost, was factored in, it would be top of the list. I think I did mention that when I covered popularity of GPUs at 4K ultra - but that is a report I only I have.
Other thing worth considering... Mainly when buying a S/H GPU. Really is if it's been mined on (less chance of that now) but also availability. The 1660S wasn't a popular gaming GPU (it's equal to a 1070) so you're more likely to find a 1080ti for more of a bargain...
Excellent video and excellent tool! Very interesting to see that Zwift hub had pretty weak connectivity. In the logs, there's no distinction between kickr bike v1, v2 and shift? I run most of my stuff on switched cabled network (cable connected direct connect to kickr v5 in race mode). Only items on wireless are my HRM (BT), Play controllers (BT) and Companion App, on (5GHz WIFI) rock stable, lowest possible latency and jitter. 4K Ultra on 65" curved 4K TV. Ran my logfile in Zwiftalizer just to confirm everything was optimized 🙂
Thanks for watching. As for distinction between kickr bike v1, v2 and shift - not that I can make out - there are a few hex values after the device name, which probably helps in environments where there's multiple of the same device - like gyms, and they are all unique, so I have to strip them out to make the aggregate counts. It's possible the version is in that string, but I don't know how to extract it yet. Yeah, the Zwift hub connection reliability isn't the best according to that report, but there are many worse. I should have included more. Maybe I'll do a follow up.
I tried to sign up, but had to reboot for non related reasons after being sent the confirmation code, but before I could enter it. Now I am unable to sign up again or do anything with the same email I used. I am not sure how else to contact you so I am writing here for assistance. Any suggestions would be greatly appreciated.
Both the tool and I have failed then. What topic would be most helpful to compare your results with the 'max' - graphics, networking, Bluetooth, ANT+, or something else entirely? Here's a summary of what most people consider best balance of max performance and budget. 1. < 15% ANT+ interference (RxFails) 2. Zero ANT+ dropouts 3. Zero Bluetooth dropouts 4. Zero TCP Disconnects 5. Zero latency test failures 6. 60 FPS average at Ultra Graphics Profile and 1440 Resolution - Throw money at 6 to increase FPS at max resolution to 4K Ultra, but a waste to buy anything above a Core i3 12100F and GTX 1660 Super or RTX 2070. - PC still the only way to get Ultra 4K, M1,M2,M3 Macs limited to High 4K and 60 FPS cap. - Neo T2 or 3 with Legacy ANT+ protocol disabled to optimize 1 and 2. - PCIe Bluetooth Card with external antenna to optimize 3. - Wahoo Direct Connect to eliminate 1,2,3 entirely assuming your local network is reliable. - Everything else depends on minimizing radio interference in your environment, reliability of you home network and stable Internet Service Provider quality.
The speed of your internet service provider does not affect your FPS. FPS is primarily determined by your computer's CPU and GPU. However, a slow internet connection can lead to lag or delays in the data being transferred, which might make the game appear to stutter, but it doesn't lower the FPS because your computer will still draw the scene with the latest data it has about riders around you, and yourself.
Speaking of waste of money: I run zwift on a 12700K and a RTX3080... but this PC is also used for other PC games and other workloads. It's a shame actually that zwift has no options to increase the graphics complexety for a higher end machine.
Nice setup. Yes, it would be nice to have more control over the graphics capabilities. I read it was on their backlog, deep deep down the list of priorities though I am guessing.
Phenomenal clarity, relevance, insights, encouragement, humour, realism, ... A must read for Zwifters than can. Much appreciated.
Excellent job. Im in the process of building/upgrading a machine. I picked up a 980ti on ebay for $79 hopefully I can get ultra 4k 60fps. Whatever I end up getting your tool will definitely come in handy. Thank you for your hard work and dedication. Donation is on its way! 👏🔥💯
Thank you! That's a good deal for a 980ti. Pair with a 12th gen Core i3 12100 or higher, or a CPU with equivalent high single core performance, to not be CPU bound in large groups. At 4K, leafy areas in Makuri Island might dip below 60FPS, but 980ti will be fine everywhere else.
Thank you for putting the effort and enthusiasm into both the tool and this video. Much appreciated 👍 Oh yes, the other runner will be along shortly I'm sure 😂
lol, thanks for watching that far in.
Excellent in depth analysis, thanks for putting all this together.
Glad it was helpful!
Excellent video and the tool itself is such a powerfull life saver.
Great details, interesting stats! Thank you Mike for taking the time to share this stuff.
Thank you for your positive feedback. I am glad you find it a useful tool.
Incredibly useful video. Thank you.
Glad it was helpful!
Fascinating video. Only came across Zwiftalizer recently when I was trying to think of how I could build/buy a PC affordably that would get me beyond Basic at average 3 fps on a good day! From your benchmarks had already deduced that an i3-121000 was probably the best buy, but what gpu? A top rate one was never going to be affordable, but after watching your video, I now don't think that I'll have to worry about this too much. Many thanks for this brilliant and instructive video.
Thanks for watching. I'm glad you found it useful.
Excellent information. Thank you
Thanks for watching
Your tool is so useful ! It's like GODMODE 🙂
It helped a lot while looking for explanations on stuttering game, and options to make it better.
I LOVE your approach on building a PC with used components for cheap !
Being in Europe I feel tricked by the 300ms ping...
Many thanks for your time building the tool and the effort doing that video. They & you are very useful.
Thank you. Your positive feedback means the world to me and is what motivates me to keep doing this.
Great video, super interesting- thank you!
One point to note is that my iPad Pro M1 runs basic profile - I think the Zwift look up table was only updated for the MacBooks for some reason!
wow, it's disappointing to hear the iPad M1 hasn't been updated. Surprising really since it's very capable.
very interesting!
thank you
A tldr would be really beneficial for these type of videos.
Good point. I made a short ruclips.net/user/shortsxZ_T1rXTtCM
Great video and thanks for the tool, I use it all the time…👍👍
Thanks for watching Neil. I appreciate your feedback and I am glad you find the tool useful.
Fascinating video; one thing that perplexes me is the affect of network latency on racing. Doesn't it mean people closer to the AWS server (people in the USA) are automatically at an advantage over the rest of the world and therefore gain an advantage in time critical aspects of the game, especially drafting. From the video there are obviously other issues that are in the mix, but how can racing and drafting be 'real' on Zwift?
No. There isn't any advantage. The server tracks each player's lag behind the central source of truth for what is "now," and modifies their time accordingly. Therefore, Zwift's in-game results are the only ones that are accurate. Other results systems that employ watchers located elsewhere will only be accurate in terms of their concept of "now" and if they account for each person's unique lag time. Pack dynamics, including position, rolling resistance, and draft, are computed client-side and sent, I believe, five times per second to the server. (What I said in the video about updating every 300 ms in not accurate - it updates every 200 ms regardless of what your lag time is). The fact that each player sees something different on their screen to where each rider is actually "now" according to the server is, in my opinion, a bigger problem. Being extremely close to the servers could actually work against you because you might believe that you were the first person across the line. In reality, though, someone who you thought was a few meters behind you might place first because the server subtracts their lag time from their finish time that you believed you were watching in real-time. In essence, it's close to real-time but never going to look accurate to everyone because of the physical distances the data has to travel are all different.
@@zwiftalizer Thanks for your reply. I think your last sentence is telling. ...if I follow you correctly racing on Zwift is a bit meh. I'm not saying it's not fun though..
@@timcorso6337 You summarised my feelings on racing accurately, and I would love to be proved wrong. It should be possible to achieve real-time gameplay with clever client-side position prediction and server time reconciliation. Maybe it's done, I just don't believe it's there yet. There's also a bunch of other factors that need to be solved for me to enjoy racing again, like sandbagging, weight doping, trainer accuracy and powermeter calibration fiddling.
@@zwiftalizer Hi, I am trying to decide whether to purchase a Zwift Ride to be used in a remote location using a mobile hotspot. I am concerned about latency with the Zwift spec being 50ms or less (my hotspot is about 150ms-175ms). In you help pages under "latency" you provide expected values of latency for places like London at ca 150ms, which I interpret as usable for Zwift but perhaps not "perfect", but contrary to Zwift-provided specs. You mention a 200ms update spec above so it would seem that a latency of 200ms or less would be compatible with a good experience on Zwift. I don't know that I would necessarily participate in Zwift racing but would like to ride through the winter season on the Zwift Ride just to keep a base level of cycling fitness/neuromuscular stimulus. I don't want to buy the Zwift Ride only to find that my 150ms latency is too jittery to have a reasonable experience on the platform. Do you have any guidance? Much appreciated and thanks for all of the work that you are doing!
@@zwiftalizer Are you saying that the discrepancy at the finish line between what we see (avatars) and the final result is a finish only phenomenon.
Really great data set - really impressed with what you have created, I havent used it for a while so I will take another look after my next session! One question - you normalized the trainer type vs dropouts for numer of session, but have you normalized for session time? i.e. dropouts per 60 min of riding? Or was the distribution of different session times across different trainers close enough such that normalization was not required? Hope that was clear!
Thanks man. That's clear. Great suggestions. I'll take another look at this report. I will factor in session time, and take something like 5th to 95th percentile dropouts to exclude outliers.
I have an old Intel i5-4590 with a Nvidia 750ti from 2015, which works ok with Zwift. But now I'm looking at upgrading my computer. The CPU will probably again be in the i5 high end. As a not gamer person I don't know/understand GPUs. So, what I would like to get from your tool -- and tried to, but not really succeeded getting the few times I looked at it in the last years -- is to get some sense of which GPU I should buy that will give me Ultra in 4K in all conditions while not be overkill (as I don't play any other game), yet still have some room for future proofness.
Intel Core i3 12100F and NVIDIA 1660 Super GPU, 16 GB RAM, NVME SSD Gen3 is the current sweet spot for Zwift in Ultra 4K in (nearly) all conditions - including large groups, in all Worlds, almost everywhere. I say 'almost' because there are some graphically complex areas in Makuri Island that need a stronger GPU, but since those spots are few and far between it does not make sense costwize to go higher than a 1660 Super. Single core performance is important in large groups. More cores are not used effectively, so there is no benefit in going up to an i5 12th generation, but doesn't hurt and will get you some room for future proofness especially if you use the PC for other things.
@@zwiftalizer Hahaha, thank you for the suggestion! I did not mean to explicitly ask you for a recommendation here, I really meant that I would like if by looking somewhere in your tool I could easily figure this out by myself. But thanks again for the answer.
@@jp_baril you make a very good point - it is not easy to figure that out from the tool. The benchmarks for Makuri (being the most complex world) is where I look, but without costs, it is not easy to know which entires are good buys and which are overkill. I only know the i3 12100F/1660 Super combo is a good balance from talking to people. It is around entry 32 in this report zwiftalizer.com/benchmarks/9/ultra/2160/100 If costs, or popularity as a proxy for cost, was factored in, it would be top of the list. I think I did mention that when I covered popularity of GPUs at 4K ultra - but that is a report I only I have.
Other thing worth considering... Mainly when buying a S/H GPU. Really is if it's been mined on (less chance of that now) but also availability. The 1660S wasn't a popular gaming GPU (it's equal to a 1070) so you're more likely to find a 1080ti for more of a bargain...
Thanks for the good work
You bet!
Do you plan to expand on this project to include other cycling sims like rouvy?
No plans to at the moment. That space is open for someone else to fill.
Excellent video and excellent tool!
Very interesting to see that Zwift hub had pretty weak connectivity. In the logs, there's no distinction between kickr bike v1, v2 and shift?
I run most of my stuff on switched cabled network (cable connected direct connect to kickr v5 in race mode). Only items on wireless are my HRM (BT), Play controllers (BT) and Companion App, on (5GHz WIFI) rock stable, lowest possible latency and jitter. 4K Ultra on 65" curved 4K TV.
Ran my logfile in Zwiftalizer just to confirm everything was optimized 🙂
Thanks for watching. As for distinction between kickr bike v1, v2 and shift - not that I can make out - there are a few hex values after the device name, which probably helps in environments where there's multiple of the same device - like gyms, and they are all unique, so I have to strip them out to make the aggregate counts. It's possible the version is in that string, but I don't know how to extract it yet. Yeah, the Zwift hub connection reliability isn't the best according to that report, but there are many worse. I should have included more. Maybe I'll do a follow up.
@@zwiftalizer maybe one reason for the discontinuing of the hub
I wasn't aware it was discontinued@@repmortskcab8483 Thanks for that update.
@@zwiftalizer only the hub classic. To be substituted by wahoo kickr core
@@repmortskcab8483 might have been a result of the legal dispute between Wahoo and Zwift
I tried to sign up, but had to reboot for non related reasons after being sent the confirmation code, but before I could enter it. Now I am unable to sign up again or do anything with the same email I used. I am not sure how else to contact you so I am writing here for assistance. Any suggestions would be greatly appreciated.
use this contact form and I'll verify your email address manually blog.zwiftalizer.com/#contact
Question remains, how do I know if my hardware Kickr Laptop is setup for max results after analyzing a ride log?
Both the tool and I have failed then. What topic would be most helpful to compare your results with the 'max' - graphics, networking, Bluetooth, ANT+, or something else entirely?
Here's a summary of what most people consider best balance of max performance and budget.
1. < 15% ANT+ interference (RxFails)
2. Zero ANT+ dropouts
3. Zero Bluetooth dropouts
4. Zero TCP Disconnects
5. Zero latency test failures
6. 60 FPS average at Ultra Graphics Profile and 1440 Resolution
- Throw money at 6 to increase FPS at max resolution to 4K Ultra, but a waste to buy anything above a Core i3 12100F and GTX 1660 Super or RTX 2070.
- PC still the only way to get Ultra 4K, M1,M2,M3 Macs limited to High 4K and 60 FPS cap.
- Neo T2 or 3 with Legacy ANT+ protocol disabled to optimize 1 and 2.
- PCIe Bluetooth Card with external antenna to optimize 3.
- Wahoo Direct Connect to eliminate 1,2,3 entirely assuming your local network is reliable.
- Everything else depends on minimizing radio interference in your environment, reliability of you home network and stable Internet Service Provider quality.
Does internet service provider speed FPS? DSL 10/1 speed here
The speed of your internet service provider does not affect your FPS. FPS is primarily determined by your computer's CPU and GPU. However, a slow internet connection can lead to lag or delays in the data being transferred, which might make the game appear to stutter, but it doesn't lower the FPS because your computer will still draw the scene with the latest data it has about riders around you, and yourself.
Tried to sign up. Says "exceeded daily email limit". ??
This is fixed. Thank you for your patience.
I wish they made the ground textures like gravel and asphalt better, that's what you look at like 80% of the time anyway, and it looks really crappy.
good video, i can´t find my logs. 😞
What type of computer are you using and where have you looked so far?
Speaking of waste of money: I run zwift on a 12700K and a RTX3080... but this PC is also used for other PC games and other workloads. It's a shame actually that zwift has no options to increase the graphics complexety for a higher end machine.
Nice setup. Yes, it would be nice to have more control over the graphics capabilities. I read it was on their backlog, deep deep down the list of priorities though I am guessing.
good video, i can´t find my logs. 😞