PassMark Software

There has been around 5 significant benchmark algorithms updates over the last 20 years. Including moving to 64bit, using AES hardware & adding SIMD to some of the sub-tests. Overall result is scaled across releases to allow a rough comparison. While there has been significant improvement in a few narrow areas (like AES encryption) overall single threaded general compute has processed only slowly. This also (partially) explains why new computer don't feel dramatically faster and old computers don't feel slow, in general desktop usage.

"Here's 20 years of data we collected." Is the software updated every year to benefit from the new hardware or is this always the same software? Processors have kept getting wider and bigger, but you need to use this silicon. I give talks on this topic: lemire.github.io/talks/2025/sea…

Introducing the USB-C Short Circuit Tester! Compact, practical, and delivers rapid testing of port voltages and short circuits— under 1 second per port — perfect for fast production line checks. For more information or to order visit: passmark.com/products/usb-s…

@Olrak29_ With a sensible naming scheme there would a lot less need to visit review & benchmark sites. So we fully support this naming madness. cpubenchmark.net/CPU_mega_page.…

Very consistent naming scheme 🤡

Pulling apart USB Type-C cables today trying to work out why some models have errors at 20Gb and some don't. We suspect the (lack of) twist in the data lines near the termination give this one occasional errors.

Windows 10 gains 1% market share back off Window 11 pcbenchmarks.net/os-marketshare…

@exQUIZitely You might have heard of Defraging, but what about Fraging? passmark.com/products/fragg…

Some of you will be old enough to remember this. Defragging - and it actually did serve a purpose. It was essential for optimizing slow, mechanical hard drives by rearranging scattered data into contiguous blocks. By doing so increased data load times (since the physical head of the hard drive didn't have to move that much). However, what I remember most was the almost hypnotical satisfaction of seeing all those little blocks flashing, being organized, and just "knowing" that it was good for my computer. In times when you fiddled around with autoxec.bat and config.sys, when every little Kilobyte of RAM mattered, and when hard drives were measured in Megabytes, not Terrabytes, the weekly routine of defragging almost felt like cleaning up your room.

Multi-threaded CPU. Currently the 356th Faster CPU available. But reasonable for it's power consumption. cpubenchmark.net/multithread/

Early single threaded benchmarks for Intel's new mobile Panther Lake chips. X9 388H & X7 368H, etc.. cpubenchmark.net/single-thread/

New version of Memtest86 was released today. Version 11.6 memtest86.com/whats-new.html Significant changes include support for new CPUs models, ECC reporting improvements, support for new RUST based UEFI coming from Microsoft and an option to treat corrected ECC errors as a failure.

@Olrak29_ Drive speeds have gotten worse since 2018. For most SSD they are still at around 120MB/sec for random reads (which is what many apps do). Need to get a, now rather rare, Optane from if you want better performance.

I genuinely thought that getting a gen 4 SSD will help in app load times, but nope. Some apps are actual dogshit

@popovicu94 My understanding is that Netflix implemented their own version of BSD Networking stack (for their specific hardware & use case). We found the out of the box network settings not very good and documentation for tweaking settings pretty poor. Like it was a hobby operating system.

@PassMarkInc Oh yeah I’ve seen a lot of networking mentions, but I can’t say I’ve looked at the data too much. I’ve also heard Linux is competitive now as well. But then again, doesn’t Netflix use BSDs extensively exactly for networking purposes?

Genuine question: What makes OpenBSD so secure? I see some claims that the code reviews are rigorous, for example. But who conducts these code reviews and what's special about the process? It's undeniable that the OpenBSD project has produced good things beyond its immediate scope, like OpenSSH is used in a lot of places now, but what's exactly the secret sauce there?

Some comments on the Ryzen 7 9700X3D CPU briefly listed on our site. For now this CPU doesn't exist and it might never exist. Our benchmarks are mostly community sourced, an average of real world performance for millions of machines. From time to time we see people trying to fake results. Sometimes for commercial gain (e.g. to win a tender, or dump on the competition), sometimes for bragging rights, media attention, for entertainment, by accident or for re-badging & selling fake parts. This has been going on for decades & isn't new. We are only a small company with limited resources, but we try and keep the public results as clean as possible. There is some automated checking and we also manually clean up the mess when required. But none the less, only around 0.1% of submissions need deletion. So generally it isn't a huge problem. It seems some media companies have setup automated scraping of our web site. So any new CPU listing gets picked up within minutes (often before we have even seen it ourselves). Any new CPU listing on our site are automatically flagged with a "high margin for error" warning until a large number of samples are received. This was the case for the 9700X3D. But this seems to be somewhat ignored in the rush to get an exclusive story. In the case of the AMD 9700X3D, the CPU was not listed on the AMD web site. So that should be suspicious, even though sometimes we do get custom or pre-production CPUs appearing. So it isn't conclusive. People also created fake photos of the non-existent CPU, which of course doesn't help when trying to determine if it is real or not. I should also state that no one in any of the media companies contacted us before or after running with the story. We would have checked the situation had someone asked us about it, but no one did. In this case the fake benchmark was done on Linux, which because of it's open source nature makes faking some details easier (compared to Windows, locked down Android & iPhones). Having said that we should also be doing a better job of validating the CPU hardware details on Linux (we'll be doing that this coming week). We had already started on overhauling the Linux release a few months ago to bring it up to the same level as the Windows release for the future V12 release of the benchmark. Ultimately some low level amount of faking & manipulation will likely continue however, as software can always emulate hardware (like in a Virtual machine (VM)). So be suspicious of new results with low submission counts. Especially if the part is new and isn't listed on the vendor's web site. We also appreciate when those in the community get into contact, letting us know when we have stuffed up, as it helps gets things corrected faster (again we are only a very small team). Was also nice of the reddit user (/A_Canadian_boi) to also come clean about their fake. Making the detection easy in this case. 9700X3D RIP

AMD’s unreleased Ryzen 7 9700X3D appears on PassMark, hinting at a lower-cost alternative to the 9800X3D for gamers. tomshardware.com/pc-components/…

The reason we know Radiation causes bit-flips in DRAM is pretty hilarious. In the late 70s, Intel Ram was occasionally producing soft, uncorrectable errors. Turns out, the ceramic packaging on the chip itself had a little bit of Uranium. You know, as one does.

@eevblog Our office is in this photo. Should have dropped in :-)

Giving a talk at a hardware meetup. They have pizza.

I actually got out of the lab and into the big smoke. Was early so I found a tree lined street to walk down.

@Pirat_Nation I want it to be true. But Win7's share is now a rounding error. pcbenchmarks.net/os-marketshare…

Windows 7 is back.

Apple's new M5 CPU appeared in the CPU benchmark charts today. Early results show extraordinarily good single threaded results and reasonable multi-threaded. cpubenchmark.net/single-thread/ cpubenchmark.net/multithread/

@BrianRoemmele We've got different stats from our users. <1% on Win7 1% on Win Server 2022 19% on Win10 (and in steep decline) 79% on Win11 But our users are typically tech savvy people pcbenchmarks.net/os-marketshare…

This month Windows 10 is no longer supported–this is 40% of Windows users. And 43% are not equipped to upgrade to the latest operating system due to system requirements. Microsoft mandates.