New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Add FIO filter? (4k, 64k, 512k, 1m separately)
In the Advanced Search?
https://serververify.com/advanced-search
Absolutely.
Reviewing how to make this work now.
Is this what you had in mind?
That's right. it will help people who need IO requirements
Ok we are working on that.
@jbiloh
I've read this multiple times and still can't figure out why it's a 4-star rating.
Why not just let reviewers provide the rating themselves instead of having AI infer it? Inferring the rating makes sense for historical reviews where no rating exists, but I am not convinced it is the best approach going forward.
It probably got tripped up by the word "feels" and that brought down its assessment of how the user actually felt about it.
I think it's ridiculous that a star rating would be given based on the text of a review.
Me: No, this isn't a 4-star review. It's definitely 5 stars.
AI: You're absolutely right! My mistake. I had no idea what the review meant - I was just guessing until a human corrected me.
This is a perfect example of a misuse of AI. They're great for summarizing things. I'd trust an AI summary for a list of features that people praise or a list of quirks that annoy them, but not trying to extrapolate more information than was provided, especially in a diverse market with competing priorities.
The review ranking was 4.4, which was previously reflected as 4 stars. That has been changed to 4.5 stars.
If you think that's not accurate report the review in the system and it'll be further looked at.
FIO/Storage Performance has been rolled out to production on Search.
1 feature that i think would be really useful (at least for me) is country level search. Sometimes I just want to find providers that have locations in a specific country, regardless of the city.
I think the current city/location search should still work as it does now (e.g. searching for "Munich, Germany" would only return providers with a location in Munich).
But if I search for a country like "Germany", it would be great if it matched all locations in that country and returned providers with locations anywhere in Germany.
Right now, if I want a list of every provider with a location in Germany, I have to select every German location from the dropdown. If I simply click "Add Germany", it doesn't return any results.
How is RAM calculated in the benchmark score/value index? It doesn't seem to be weighted very strongly, if at all?
I feel like it shouldn't impact the "benchmark score" (it wasn't benchmarked so makes sense) and then instead be bucketed based on RAM capacity.
Something like:
So that a 64G server can still be a good value within its "weight class" as it were, and have that be visible with the value index not being some incomparable single digit value.
Same argument could be made for storage capacity in a "storage server" bracket.
Great suggestion.
Fully implemented, and in record time!
That is a fair point. RAM capacity does not affect the benchmark score itself, since that score is based only on measured performance. It is considered in the value index, but its weighting is intentionally limited so that a large amount of inexpensive RAM does not overpower poor CPU, disk, or network performance.
I do like the broader idea of making it easier to compare servers within similar resource classes. Rather than creating quite that many fixed buckets, we may be able to add broader RAM and storage capacity ranges as filters or separate value comparisons. That would allow a 64 GB server to be evaluated against comparable 64 GB-class servers without making the main value index difficult to understand or fragmenting the results too heavily.
We will give this some thought as we continue refining the value index and Advanced Search. Thanks for the detailed suggestion.
Consider my complex approach.
This feature meant the user could mark plans above the minimum spec. Sorting by price often yields heterogenous specs. Instead, I think it's better to do pareto approach: increase the budget and get more bang for your bucks
I'm spending some time thinking through this over the next few days.
Any other suggestions or features to make the system more useful?
Not really. But I'd suggest looking at openrouter's implementation when comparing price/performance (based on several category) and the pareto feature