Banco de pruebas Hardlimit

[krampak]

@cobito when you click to sort the results by Multi mode, it always sorts by Single mode. I wanted to see which are the fastest CPUs I've validated when sorting by Multi mode.

https://bm.hardlimit.com/search.php?s=krampak&o=1

[cobito]

@krampak It's already fixed: mono and multi were swapped.

[cobito]

Changes have been made that provide significant improvements to the loading times of all pages. We have also corrected some text formatting that was incompatible with the new version of the language selector.

Additionally, a few days ago we began marking results from overclocked CPUs in the list of latest cover page submissions.

[cobito]

Single-thread and multi-thread per-user rankings are now displayed on the homepage. The ranking shows the best result per user-CPU to prevent a single user from dominating multiple positions with the same configuration.

[cobito]

The first beta of the new version is now available. Since the executable is unsigned, you'll need to disable Smart App Control to test it. Be careful with this, because if you disable it globally on recent versions of Windows 11, it cannot be re-enabled without reinstalling the OS. I'm not sure if it's possible to add an exception for a specific executable.

In essence (and without going into the details of internal changes), a new "Monitor" tab has been added, which displays a graphical representation of total CPU usage, the most loaded core (the one with the highest percentage usage), the current frequency, and memory usage.

The button to show more CPU information has been fixed. Now, clicking it opens the processor spec sheet in the browser.
An artificial limitation has also been found that prevents detecting more than 96 threads, which explains some results from monstrous CPUs that were submitted and where this limit was visible. This remains unresolved for now.

The download is available here.

With this version, you can validate results as if it were the current stable version.

[pos_yo]

@cobito I took a quick look at it (I tested it with Wine from Linux) and it looks more modern, and the monitor seems very useful to me... I ran the test to compare it with the one I did on W11 and... well... I already figured it would have lower performance...

[krampak]

@cobito Tested, everything seems fine at first. It did take a slight hit on performance, but I also have my entire development environment loaded and I haven't rebooted in days. If you need me to, I can test it again. I'm not sure if it did this before, but I like that it compares it to the registered average for your processor. It saved me from having to look up old results.

As for the resource monitor, I notice it updates very slowly while the test is running, but at idle the graph updates extremely quickly, as if the elapsed time axis weren't the same. Is this normal?

[_Neptunno_]

@cobito it wasn't the first time I took the test, but it was the first time I registered it with my nickname (at least I think so). The thing is, I noticed a slight discrepancy between what the test itself says and what appears on the page. Well, just to be clear, it doesn't really matter to me, I'm just mentioning it in case it helps or if that's just how it works:

As always, excellent work

Cheers!!

[cobito]

@krampak: Regarding the resource monitor, I notice it moves very slowly when the test is running, but while idle the graph runs at full speed, as if the elapsed time axis weren't the same. Is this normal?

The design intention for the graphs is for them to refresh quickly, much faster than in Task Manager. At the moment (this isn't the final number), the refresh rate is 24Hz.

It's normal for them to lag during the test because hlbm-core hogs all the CPU time on that thread. During the test, the graphs shouldn't be displayed since they consume a small amount of CPU, which alters the result (that might be why your score dropped). The end goal is to prevent tab switching during the test and instead display numerical consumption values on the "Benchmark" tab with a lower refresh rate.

@_Neptunno_:

@cobito it wasn't the first time I've taken the test, but it was the first time I've logged it under my username (at least I believe so). Anyway, I noticed a slight discrepancy between what the test itself reports and what shows up on the website. Well, just to be clear, I don't mind at all; I'm just pointing it out in case it helps or in case it's actually supposed to work this way:

Indeed, there is a discrepancy in methodologies that was accidentally introduced in v.4.x.x of the central server. Actually, it's a bug on the server side, not in the program. Thanks for the reminder, as I don't think I even had it noted down.

Thanks to everyone for the comments and for running the tests.

[cobito]

Central

The "process" (manufacturing process) field has been added, which now shows both the commercial name and the gate width on the processor page (cpu.php).

Program

• @pos_yo now appears in the credits in the "About" tab as a beta tester.
• A stress test has been added (Stress Test tab). You can choose single-thread or multi-thread mode. There are two modes that cannot be configured by the user:
  • "Normal" (AVX2 + FMA 3) for Haswell/Steamroller and newer.
  • "Compatibility" (SSE4.1) for current low-cost processors and from the Penryn/Bulldozer generation onwards. For processors older than those, stress test support is not provided unless there is interest.
• During the stress test execution, only the Stress Test and Monitor tabs can be viewed.
• Software requirements are now 64-bit Windows 10 and 11. It is not compatible with 32-bit systems or Windows 7 and earlier.
• Now, during benchmark execution, you cannot switch tabs to avoid additional overhead during the tests.
• During benchmark execution, total CPU usage and the usage of the most loaded core are displayed in numerical values (at a 1Hz update rate).
• The brand and part number of each memory module are now displayed next to its slot number.
• A button has been added to view the details available in the central database for the memory installed in each slot.
• Clicking the logo in the "About" tab now opens the forum.
• There are no changes related to the test bench Core: you can continue validating as usual, and there should be no behavioral changes in this regard.

You can download this build from here.

[krampak]

@cobito said:

Central

The "process" (manufacturing process) field has been added, showing both the commercial name and the gate width on the processor sheet (cpu.php).

Program

• @pos_yo now appears in the credits on the "About" tab as a beta tester.
• A stress test has been added (Stress Test tab). You can choose single-thread or multi-thread mode. There are two modes that cannot be configured by the user:
  • "Normal" (AVX2 + FMA 3) for Haswell/Steamroller and later.
  • "Compatibility" (SSE4.1) for current low-cost processors and from the Penryn/Bulldozer generation onwards. For processors older than those, stress test support is not provided unless you have an interest in it.
• During the stress test execution, only the Stress Test and Monitor tabs can be viewed.
• Software requirements are now Windows 10 and 11 (64-bit). It is not compatible with 32-bit systems or Windows 7 and earlier.
• Now, during benchmark execution, you cannot switch tabs to avoid extra overhead during the tests.
• During benchmark execution, total CPU usage and the most loaded core's usage are displayed in numerical values (1Hz refresh rate).
• The brand and part number of each memory module are now displayed next to its slot number.
• A button has been added to view the details available in the central database for the memory installed in each slot.
• Clicking on the logo in the "About" tab now opens the forum.
• There are no changes related to the Core of the benchmarking suite: you can continue validating as usual, and there should be no behavioral changes in this regard.

You can download this build from here.

As a suggestion, I would say try to match the design across the different tabs. I would aim to always center the content (both vertically and horizontally), as right now everything is a bit mixed up, and now that you've increased the form size, it's more noticeable (I assume it was for the Monitor section).
By the way, does the end of the monitor's graph (left side) do weird things? It looks like a whip

[cobito]

@krampak Thanks for the notes. Noted.

I'll sort out the aesthetics once everything is more clearly defined. I enlarged the window for the Monitor, yes, but also to display more information in pre-existing tabs. For example, the memory tab now shows up to 8 memory slots that take up practically the entire area when they are present.

Yes, the graph does some weird things. It's due to the moving average. The statistics show too much variability if displayed at the refresh rate, and honestly, it doesn't produce a graph that's easy/pleasant to interpret. That's why it's smoothed using a moving average, which is what fails when the samples start dying out. Noted.