Hardlimit test bank

[Magog]

I have put in an 8700k. Still untouched, all by default.

Best regards.

[cobito]

@magog Perfect. Having data from micros to stock frequency (in addition to overclocked) is important in order to be able to extract future overclocking statistics.

By the way, if everything goes well, version 1.2 of the test bench will appear this weekend.

[Magog]

Soon I will upload more with OC done. By the way, how has this changed!! I'm used to the BIOS of the 1366, this one has too many frills.
And on the other hand, now that the change has been made, I can only reaffirm that the socket 1366 with the Xeon was and is an excellent platform.

Best regards.

[cobito]

Version 1.2 is now available. For all the details, you can visit the first post of this thread. The most notable is the deep aesthetic change, but there are also some important changes under the hood: after analyzing many results, it was concluded that modes 1, 2, and 3 corresponding to MMX, SSE, and SSE2 make no sense today, so they have been removed. In this way, SSE3, AVX, and AVX2 modes are now modes 1, 2, and 3 respectively (the results of the old modes 1, 2, and 3 are still available for consultation, but they are no longer used to form the statistics).

Now the Windows version is also sent, although for the moment this information is not reflected in the Central. I am going to give it a good debugging soon, so I will take the opportunity to make the Windows version come out.

To finish, tests 3 (memory) and 4 (generic ops) have been swapped. This change also affects all past results. The tests are now:
· Test#1: Integers
· Test#2: Floating point
· Test#3: generic operations (zip de/compression)
· Test#4: memory operation performance

As usual, bug reports and suggestions are welcome.

Download

[Namiga]

I'm testing. Bit Defender keeps insisting that there is malware
What a cross... what a fuss this antivirus is...

[cobito]

@namiga It's inevitable. I don't know if any of you know if signing the executable would prevent these things. I will try to get in touch with the main antivirus developers so that they can whitelist the test bank.

Edited: in virus total it gives 7/62 positives. Honestly, what a job...

[cobito]

@krampak has already fixed the "identification problem" of your i3-8350k and it shows all the information correctly. It was actually a bug in the algorithm that finds similarities between processors and also affected other models.

[krampak]

What a facelift the new version is!! Bravo.

Edit: @cobito It seems that something new is working well. Nothing is updated, not the Current Trask, not the progress bars, not even when the test ends (after the final hlbm-detect).

[cobito]

@krampak Has that happened to you multiple times? Has it happened on multiple PCs?

[Magog]

To me without problems.

Best regards.

[krampak]

@cobito I've only tested it on my home PC. It was working fine now (running as Administrator), so I went to test running it without forcing administrator and when I closed the benchmark for the X, it continued running the tests in the background, so when I reopened the benchmark (correctly) and clicked Start, it closed abruptly without giving any message.

Edit: Now the current task and the bars look good, whether or not you force administrator mode.

[cobito]

@krampak This new version includes some extra security measures to prevent tampering with results. One of them is to prevent multiple instances of the test bank from running simultaneously, which is why you were closed without explanation when there were other 'core' processes running (in general, the policy is not to offer any information so as not to give clues, but perhaps I should make a warning message appear for these accidental cases, which I think are going to be quite common).

Regarding what happened to you the first time, it seems to have been an internal interface error, although I don't know what the cause could be. If it repeats, comment on it again and I'll send you a version with a debugging window to know exactly what is happening.

Regarding the fact that the processes continue to run after closing the window, this is the expected behavior until a mechanism is included to cancel the test. When this happens, all you have to do is wait a couple of minutes for the test bank to finish. After that, the processes will close automatically and you can run the program again. The cancel button was going to be included in this version, but with so many changes, I decided to release this version first, to get all the cases, and then add new functionalities.

[QVENGADOR]

Hackers on the web? That's what I wondered when I received so many emails about test bank. The thing is, I thought it was spam, but I can see now that it's not.
Which CSM do you use for the web?

[cobito]

@qvengador The mail service has been down for a few weeks. When we fixed it, it seems like it sent everything it had in the queue all at once.

[krampak]

I just tested the new version on a PC with Windows 10 and defender and I got this warning @cobito

[cobito]

@krampak Thanks, it's already been reported. For now, McAfee has already fixed the false positive. I hope that in the coming days the rest will be resolved.

[cobito]

Microsoft has given the green light to the file. If anyone has Windows Defender with definition 1.259.206.0 or higher and can confirm it to me, I will be more at ease.

It can happen that some antivirus gives a positive result, but for Windows Defender to wipe it out at the first opportunity gives a sorry image.

[cobito]

Now only results obtained from stock frequency validations are calculated. In return, a overclocking ranking appears on the tab of each model. Example.

The system now calibrates the frequencies sent by the test bench but for this it has to learn based on the results sent. The more results are sent at stock frequency, the better the measured frequency will be. For the moment this only applies to the tab of each model and not to the individual results.

As always, any report of failure or strange thing will be welcome.

[krampak]

@cobito said in Hardlimit test bank:

Microsoft has given the green light to the file. If someone has Windows Defender with definition 1.259.206.0 or higher and can confirm it to me, I'll be more at ease.

It can happen that some antivirus gives a positive, but that Windows Defender wipes it out at the first opportunity gives a pitiful image.

Confirmed, I didn't get the warning this time.

[cobito]

Some important changes in the Central.

Now the overclocked validations have a different treatment than the stock frequency validations. That makes the models for which there are no stock frequency results or whose measured frequencies show inadequate test bench conditions* not appear in the statistics. For example, there is the i7-6959X from @Xevipiu that no longer appears in the cover ranking or in the comparisons because there are no stock frequency results. In addition, the datasheet does not show all the information or this is inconsistent.

On the other hand, now the frequency shown in the results (page that appears when validating) is corrected based on the previously sent results. As I said above, the more results there are at stock frequency, the more precise and accurate the frequencies shown will be. In principle, with 3 or 4 validations in any mode, it should be enough to show an adequate frequency.

Regarding the results of overclocked micros, as I also mentioned above, in each datasheet there is a ranking of individual results showing the user, the frequency, the scores and the motherboard model. Example.

*The frequency measurement is done on a single core, so if there are background programs hogging part of the processor and consuming several cores, the measured frequency will not be the "boost" frequency but the one provided at medium load or at full load. If there is a model in which in all results this interference has occurred, this model will not appear in the statistics.

Regarding the problems of these changes: the back-end has become very slow (again). In the next code sweep, optimization will be the priority.

EDITED: The performance problems and some minor bugs on certain pages have been fixed.