News Memtest86+ Is Back! New Version Released After 9 Years

[Admin]

Out of nowhere, Memtest86+ has received its first major up since 2013, and is now in active development. The new upgrade brings the application up to speed, supporting hardware from 2014 all the way up to the modern era.

Memtest86+ Is Back! New Version Released After 9 Years : Read more

 

[tennis2]

What's better now then? The free version of Memtest86 v10 or Memtest86+ v6?

Or are they both worthless.

 

[setx]

Commercial Memtest86 is definitely worthless. But how this new Memtest86+ compares to TestMem5 is actually interesting.

 

[josmat]

Just tried downloading this new Memtest86+. First downloaded the Windows USB Installer. On VirusTotal 7 antivirus flagged the file as malicious. Then I downloaded the ISO file, and Windows Defender detected a Trojan:Script/Wacatac.B!ml.

I know it's a very low level tool that could trigger false positives. But it's suspicious anyway. I'll wait for further community approval.

 

[bit_user]

Commercial Memtest86 is definitely worthless.

Why do you say that?

 

[InvalidError]

MemTest86+ 5.31 was released in 2020 and I'm almost certain there were some point-releases between that and 2013, though most of memtest86's history appears to have been scrubbed from the site or is tucked away somewhere I couldn't easily locate it.

 

[jkhoward]

MemTest86 algorithm is almost certainly being used by PassMark in their memory benchmarks in some capacity.

 

[PassMark]

Some background.

MemTest86 (without the plus) was originally developed by Chris Brady (BradyTech Inc) with a first release in 1994. However, some of the testing algorithms used have been under development since 1981 and were previously implemented on Dec PDP-11, VAX-11/780 and Cray XMP (an old super computer) architectures. Since then, there have been more than a dozen new versions released. Between 1994 and 2002, support for 64bit was added (but only via PAE paging), new CPU types, symmetrical multiprocessors and many other features have been added during this period. MemTest86 was released as free open source (GPL) software during this period.

Between 2002 – 2004 (during the time period of the MemTest86 V3.0 release and there wasn't much activity from Chris) the code was 'forked' by Samuel Demeulemeester (formally part of the French CanardPC publishing group in France) into another version of the software called MemTest86+. The naming was unfortunate as it now leads to total confusion between users. Nevertheless, both Memtest86s subsequently continued to be developed. The original MemTest86 was continued to be developed across 2004 to 2013, with multiple updates made - most notably the stable Version 4.0 release. MemTest86 (the original) and the plus version exchanged a bunch of source code in this period.

In Feb 2013, after another long period of not much development, PassMark Software took over the maintenance of the original MemTest86 project from Chris. This was around the time that a lot of technological changes were occurring. The 64bit era was here, DDR4 was coming, UEFI had already arrived and Microsoft's Secure boot technology threatened to prevent MemTest86 from booting on future PC hardware.

Starting from MemTest86 v5, the code was re-written to support self-booting from the newer UEFI platform. The core software still remains free to use without restrictions. The MemTest86 v4 project (for traditional BIOS) is still maintained and remains GPL open source, for use on old machines. However, from V5 with the the software is being released under a proprietary license.

PassMark did five major releases between 2013 and today. PassMark added ECC support, Secure boot support (with Microsoft code signing it), Mouse support, a graphical UI, logging of test reports to the disk, Native 64bit, Some 128bit testing, support for a heap of new CPUs, DDR5 support, row hammer testing, command line boot configuration, PXE network booting, multi-language support, inventory checking, ECC injection, debug logging, performance improvements, ARM CPU support, memory address decoding (to tell you exactly which memory stick and chip on the stick is bad) and a bunch of other stuff.

Development on the MemTest86+ version was discontinued as there have been minimal updates since 2013 (until today that is, with this new V6 release). So the plus version was around 9 years behind. This release today brings it forward a few years. So for example, it now has DDR5 support, but is still missing a lot of important stuff like ECC support, Secure boot support, row hammer testing and the ability to log a test report to disk. The lack of secure boot being especially problematic for novice users as Win11 requires secure boot to be on, but Memtest86+ requires it to be off. So it won't work on most machines without a deep dive into the BIOS settings.

Nonetheless the core test algorithms in both versions were common for a long while and can be traced back 40 years. They are diverging slightly now but once you get either version running they should both provide a solid RAM test.

 

[alceryes]

Commercial Memtest86 is definitely worthless. But how this new Memtest86+ compares to TestMem5 is actually interesting.

Why is it worthless? Doesn't the free version (of the commercial Memtest86) still allow the full test suite up to 4 passes?
You really only need 1 pass for 99.9% stability. The last couple tests in the suite are more brutal on the RAM than just about anything else you can throw at it.

Edit - It's the USB creator part of the Windows version that several AV softwares don't like. Could be a false positive. The Linux ISO version scans clean so people can just use that and create the bootable USB using Rufus. The free version of Memtest86 (Passmark's version) is still completely adequate to determine memory stability/fault.

 

[boe rhae]

Some background.

MemTest86 (without the plus) was originally developed by Chris Brady (BradyTech Inc) with a first release in 1994. However, some of the testing algorithms used have been under development since 1981 and were previously implemented on Dec PDP-11, VAX-11/780 and Cray XMP (an old super computer) architectures. Since then, there have been more than a dozen new versions released. Between 1994 and 2002, support for 64bit was added (but only via PAE paging), new CPU types, symmetrical multiprocessors and many other features have been added during this period. MemTest86 was released as free open source (GPL) software during this period.

Between 2002 – 2004 (during the time period of the MemTest86 V3.0 release and there wasn't much activity from Chris) the code was 'forked' by Samuel Demeulemeester (formally part of the French CanardPC publishing group in France) into another version of the software called MemTest86+. The naming was unfortunate as it now leads to total confusion between users. Nevertheless, both Memtest86s subsequently continued to be developed. The original MemTest86 was continued to be developed across 2004 to 2013, with multiple updates made - most notably the stable Version 4.0 release. MemTest86 (the original) and the plus version exchanged a bunch of source code in this period.

In Feb 2013, after another long period of not much development, PassMark Software took over the maintenance of the original MemTest86 project from Chris. This was around the time that a lot of technological changes were occurring. The 64bit era was here, DDR4 was coming, UEFI had already arrived and Microsoft's Secure boot technology threatened to prevent MemTest86 from booting on future PC hardware.

Starting from MemTest86 v5, the code was re-written to support self-booting from the newer UEFI platform. The core software still remains free to use without restrictions. The MemTest86 v4 project (for traditional BIOS) is still maintained and remains GPL open source, for use on old machines. However, from V5 with the the software is being released under a proprietary license.

PassMark did five major releases between 2013 and today. PassMark added ECC support, Secure boot support (with Microsoft code signing it), Mouse support, a graphical UI, logging of test reports to the disk, Native 64bit, Some 128bit testing, support for a heap of new CPUs, DDR5 support, row hammer testing, command line boot configuration, PXE network booting, multi-language support, inventory checking, ECC injection, debug logging, performance improvements, ARM CPU support, memory address decoding (to tell you exactly which memory stick and chip on the stick is bad) and a bunch of other stuff.

Development on the MemTest86+ version was discontinued as there have been minimal updates since 2013 (until today that is, with this new V6 release). So the plus version was around 9 years behind. This release today brings it forward a few years. So for example, it now has DDR5 support, but is still missing a lot of important stuff like ECC support, Secure boot support, row hammer testing and the ability to log a test report to disk. The lack of secure boot being especially problematic for novice users as Win11 requires secure boot to be on, but Memtest86+ requires it to be off. So it won't work on most machines without a deep dive into the BIOS settings.

Nonetheless the core test algorithms in both versions were common for a long while and can be traced back 40 years. They are diverging slightly now but once you get either version running they should both provide a solid RAM test.

Fantastic explanation, thank you

 

[setx]

Why do you say that?

Why is it worthless? Doesn't the free version (of the commercial Memtest86) still allow the full test suite up to 4 passes?

I feel that overclocking community completely moved from Memtest86 to TestMem5 nowadays. I haven't done much testing, but for several tries TestMem5 found errors much faster so I didn't bother to wait would Memtest86 find the same errors in the end or not.

 

[alceryes]

I feel that overclocking community completely moved from Memtest86 to TestMem5 nowadays. I haven't done much testing, but for several tries TestMem5 found errors much faster so I didn't bother to wait would Memtest86 find the same errors in the end or not.

It looks like TestMem5 is still 32-bit and hasn't been officially updated (by it's creator) since 2010. Some of the newer forks on Github trigger Google and other AV softwares to flag them as malicious.
The probable reason for why it found errors quicker is because it jumps right into the harder memory tests first with the config you used.

I'll stick with Memtest86.

 

[in_the_loop]

", MemTest86 was originally created back in the mid 1990s, and was one of the earliest DDR memory testing applications for personal computers. "

DDR-memory in the mid 90:s?

 

[bit_user]

"... was one of the earliest DDR memory testing applications for personal computers. "

DDR-memory in the mid 90:s?

Yeah, I remember when FPM (Fast-Page Mode) DRAM came onto the scene. Then came EDO (Extended Data-Out) DRAM. After that, came DDR (sometime in the early 2000's). DDR2 hit around 2004-2005, at about the same time as the transition from AGP to PCIe.

Our family PC was a 80386 and had chip DRAM, IIRC. A whopping 8 MB, back in the days when 2-4 was common. I could be mistaken, but I think we also had SRAM cache that was implemented as chips onboard.

Then, when we got a Pentium, which definitely had SIMMs. I think probably FPM DRAM, but I'm not sure.

 

[Niva]

Yeah, I remember when FPM (Fast-Page Mode) DRAM came onto the scene. Then came EDO (Extended Data-Out) DRAM. After that, came DDR (sometime in the early 2000's). DDR2 hit around 2004-2005, at about the same time as the transition from AGP to PCIe.

Our family PC was a 80386 and had chip DRAM, IIRC. A whopping 8 MB, back in the days when 2-4 was common. I could be mistaken, but I think we also had SRAM cache that was implemented as chips onboard.

Then, when we got a Pentium, which definitely had SIMMs. I think probably FPM DRAM, but I'm not sure.

I had to look this up, first use of DDR in PCs for memory was in 2000, but for some reason I remember it being around in the mid 90s... not sure in what context. Wow I'm way too old. Cheers'yall~!

 

[mitch074]

Yeah, I remember when FPM (Fast-Page Mode) DRAM came onto the scene. Then came EDO (Extended Data-Out) DRAM. After that, came DDR (sometime in the early 2000's). DDR2 hit around 2004-2005, at about the same time as the transition from AGP to PCIe.

Our family PC was a 80386 and had chip DRAM, IIRC. A whopping 8 MB, back in the days when 2-4 was common. I could be mistaken, but I think we also had SRAM cache that was implemented as chips onboard.

Then, when we got a Pentium, which definitely had SIMMs. I think probably FPM DRAM, but I'm not sure.

Since the RAM controller was on the chipset, you could get pretty much everything on a Pentium. That said, most common were 32-bit FPM modules (by pair, because of 64-bit wide front side bus), EDO (same physical format, ditto) and SDRAM, which lasted until the Pentium 4 era. DDR being the same technology with a doubling of the transfer rate (using the descending cycle along with the raising cycle) as SDRAM the same way AGP did, technically it didn't exist yet but when it came out only Intel ignored it.

 

[mitch074]

I had to look this up, first use of DDR in PCs for memory was in 2000, but for some reason I remember it being around in the mid 90s... not sure in what context. Wow I'm way too old. Cheers'yall~!

You're remembering SDR-SDRAM, which came out around 1995. DDR-SDRAM came a few years later but was still very similar, causing some confusion at the time.

 

[bit_user]

... and SDRAM, which lasted until the Pentium 4 era. DDR being the same technology with a doubling of the transfer rate (using the descending cycle along with the raising cycle) as SDRAM the same way AGP did,

Yeah, when DDR hit the scene, I think they called it "DDR SDRAM". After a while the "SDRAM" part got dropped, in colloquial usage. BTW, the S was for "Synchronous". From Wikipedia's description, it sounds like that's referring to the use on an interface stage that's latched by a separate clock signal.

technically it didn't exist yet but when it came out only Intel ignored it.

They had that whole misadventure with RAMBUS. The only time I ever saw RAMBUS DRAM was in a workstation PC we had at work. Funny enough that was also a Pentium 4 Xeon, which was a fairly new branding, back then. Might've even been dual-CPU.

I think I remember Intel delaying integration of the memory controller into their CPUs because they wanted flexibility of whether to go with RAMBUS or DDR. Meanwhile, it was a differentiator that gave AMD's Opteron a distinct advantage. I'm surprised Intel didn't even do it in Core 2 (after RAMBUS was basically dead & gone), but the generation after.

Not sure if it's related, but I accidentally bought Registered DIMMs for my Pentium 4. Luckily, I was able to ebay them pretty easily because Opterons needed (or at least supported) RDIMMs. I wonder if it took a couple generations for AMD's integrated memory controller to start supporting Unbuffered DIMMs.

 

[bit_user]

You're remembering SDR-SDRAM, which came out around 1995. DDR-SDRAM came a few years later but was still very similar, causing some confusion at the time.

I doubt we had SDR SDRAM SIMMs/DIMMs in 1995. I know the sequence went: FPM, EDO, then SDR. I just don't remember exactly when. If it helps, here's an Intel 440BX Pentium II board, from 1998, that lists both EDO and SDR compatibility:

https://www.anandtech.com/show/138​

And also in 1998, Anand was talking about building budget PCs, when he said "For memory, standard 72-pin EDO SIMMs will work just fine."

https://www.anandtech.com/show/101/3​

Ah, and in May 1998 was his article SDRAM: Hype or the Future:

https://www.anandtech.com/show/39​

​

(I think we can safely say the latter! ; ))

 

[mitch074]

I doubt we had SDR SDRAM SIMMs/DIMMs in 1995. I know the sequence went: FPM, EDO, then SDR. I just don't remember exactly when. If it helps, here's an Intel 440BX Pentium II board, from 1998, that lists both EDO and SDR compatibility:

https://www.anandtech.com/show/138​

And also in 1998, Anand was talking about building budget PCs, when he said "For memory, standard 72-pin EDO SIMMs will work just fine."

https://www.anandtech.com/show/101/3​

Ah, and in May 1998 was his article SDRAM: Hype or the Future:

https://www.anandtech.com/show/39​

​

(I think we can safely say the latter! ; ))

SDRAM came out in 1993. By end of 1995, it was entering the consumer market - Via's Apollo VP chipset had SDRAM support and was announced in Q4 2015 (availability : Q1 1996), Ali's Aladdin III chipset : 1996 too. Intel was late to the party and only added SDRAM support in 1997.
So SDRAM wasn't yet the dominant RAM format, and Intel tried to kill it off with Rambus for the Pentium 4 generation, but AMD's Athlon kept it kicking until it was replaced with first-get DDR SDRAM - Via, Sis and Ali all put out chipsets (both for AMD and Intel processors, that had separate sockets by then) that supported it until Intel dropped Rambus and got into DDR too, starting in 2002.

 

[mitch074]

Yeah, when DDR hit the scene, I think they called it "DDR SDRAM". After a while the "SDRAM" part got dropped, in colloquial usage. BTW, the S was for "Synchronous". From Wikipedia's description, it sounds like that's referring to the use on an interface stage that's latched by a separate clock signal.


They had that whole misadventure with RAMBUS. The only time I ever saw RAMBUS DRAM was in a workstation PC we had at work. Funny enough that was also a Pentium 4 Xeon, which was a fairly new branding, back then. Might've even been dual-CPU.

I think I remember Intel delaying integration of the memory controller into their CPUs because they wanted flexibility of whether to go with RAMBUS or DDR. Meanwhile, it was a differentiator that gave AMD's Opteron a distinct advantage. I'm surprised Intel didn't even do it in Core 2 (after RAMBUS was basically dead & gone), but the generation after.

Not sure if it's related, but I accidentally bought Registered DIMMs for my Pentium 4. Luckily, I was able to ebay them pretty easily because Opterons needed (or at least supported) RDIMMs. I wonder if it took a couple generations for AMD's integrated memory controller to start supporting Unbuffered DIMMs.

Intel delayed including an IMC because they didn't want to change sockets just yet - Core and Core 2 both slotted into the latter Pentium 4 socket (chipset compatibility was another matter : you actually had to look at your chipset's REVISION, not only model, to know if you could use a Core processor in your P4 mobo).
As for Opteron requiring registered RAM, at the time AMD was still tinkering with the IMC, and their first dual channel implementations; they had a socket for Opteron (940), a socket for Athlon64 X2 (939) and another for other chips (754) - if memory serves me. They must have waited for DDR2 to come out and standardize, because DDR was getting long in the tooth by then.

 

[dandano]

Does anyone know if you can turn off the large PASS/FAIL messages that display over the address messages? I can't see where the failures are.

 

[bit_user]

Does anyone know if you can turn off the large PASS/FAIL messages that display over the address messages? I can't see where the failures are.

It's been a while since I ran it, but isn't there an option to save a report at the end? If so, I think that should have the addresses.