The reason, why some chips overclock so high is date of manufacture, not code (part number) of the memory chips.
So knowing actual part number, for example HY57V28820HCT-H, is not crucial how high average overclock on these chips will be, but date of manufacture.

When you check all your V) category chips (Extreme), those are made from about august 2006 to year 2007... and one memory stick is even made from chips made in 2009!
In that time, the bad chip with zero overclock from 133 MHz just couldn't be made, because same process was already enhanced to maximum and set for late DDR1 memories. They were produced in that time for various OEM and industrial machines, so there are replacement units available.
Even despite the fact, that they could already make PC3200 DDR1 memories, they just made chips same way, set them internally as SDR instead DDR, and marked them by highest official number: PC133. Because that was standard, and that was according what customers ordering replacement units were orienting. In other words, despite the fact, that most of the chips made in 2006 and later could reach 166-200 MHz , sometimes even more, they've marked them PC133 anyway, so ordering lists are not out of stock. They didn't marked them PC150 or PC166, because that could confuse customers and it could lead to situation, that they will pick competition for PC133 replacement units.

So PC150 or PC166 has only historical importance, as for example in year 2000 or 2001, most of the chips could not reach 150 or 166 MHz frequency, but when you bought expensive PC150 or PC166 in that time, you had guarantee thay are rated for this speed . They were cherry-picked in that time.

Of course now is the trick to pick PC133 memory module with chips from quality manufacturer like Samsung and try to get some 2006-2009 -ish modules. There is great chance, you will end up with some that are able to achieve insane overclocks close to 200 MHz or over 200 MHz. But in 2000 or 2001, there was no other choice as to pick PC150 or PC166 memories. With PC133, you were not usually lucky, and for sure, you couldn't reach 180-250 MHz speeds in that time, even if you would bought 1000 of them.

W.x. wrote on 2025-10-20, 23:53:

The reason, why some chips overclock so high is date of manufacture, not code (part number) of the memory chips. So knowing actu […]
Show full quote

The reason, why some chips overclock so high is date of manufacture, not code (part number) of the memory chips.
So knowing actual part number, for example HY57V28820HCT-H, is not crucial how high average overclock on these chips will be, but date of manufacture.

When you check all your V) category chips (Extreme), those are made from about august 2006 to year 2007... and one memory stick is even made from chips made in 2009!
In that time, the bad chip with zero overclock from 133 MHz just couldn't be made, because same process was already enhanced to maximum and set for late DDR1 memories. They were produced in that time for various OEM and industrial machines, so there are replacement units available.
Even despite the fact, that they could already make PC3200 DDR1 memories, they just made chips same way, set them internally as SDR instead DDR, and marked them by highest official number: PC133. Because that was standard, and that was according what customers ordering replacement units were orienting. In other words, despite the fact, that most of the chips made in 2006 and later could reach 166-200 MHz , sometimes even more, they've marked them PC133 anyway, so ordering lists are not out of stock. They didn't marked them PC150 or PC166, because that could confuse customers and it could lead to situation, that they will pick competition for PC133 replacement units.

So PC150 or PC166 has only historical importance, as for example in year 2000 or 2001, most of the chips could not reach 150 or 166 MHz frequency, but when you bought expensive PC150 or PC166 in that time, you had guarantee thay are rated for this speed . They were cherry-picked in that time.

Of course now is the trick to pick PC133 memory module with chips from quality manufacturer like Samsung and try to get some 2006-2009 -ish modules. There is great chance, you will end up with some that are able to achieve insane overclocks close to 200 MHz or over 200 MHz. But in 2000 or 2001, there was no other choice as to pick PC150 or PC166 memories. With PC133, you were not usually lucky, and for sure, you couldn't reach 180-250 MHz speeds in that time, even if you would bought 1000 of them.

That era was "weird"

Yes you have a point about chip production dates
But the example you used is not a proper one.

The sdram chips and their speed and latencies were improved , not from a better manufacturing process and newer production dates of the same ram chip type .... But from newer ram chips models.
The need for faster ram was initiated by the production of newer board chipsets (Pentium 4 - sdram era) that required different specs.

Also keep in mind that industries were not bining their ram chips productions back then ... and most of the ram manufacturers didn't also.

Their main goal back then was to meet demand , produce and sell PC100 AND PC133 spec ram modules.
Because that was the market.

So,many of their production range could overcome the PC133 standard in real life.

Also , voltage spec of that era was up to 3,6 volt max from the factory. (in some cases)

There were very few ram overclocking oriented companies back then , that's why the PC150 and PC166 modules were not that "popular" ... and now we call them "rare modules" cause of the limited production for the "enthusiast".

Too many of the "ordinary" 7.5ns chips from Samsung and Hyundai (later Hynix) could easily reach PC150 speed , and it didn't have to be hand-picked.
Of course the "hand-picked" marketing also worked at that time , and there were companies that really did so , with very good results. (mushkin,corsair,diamond,ocz)

The 7ns ram chips was the "move ahead" of that era , followed by the real 6ns ram chips.
But also keep in mind that , the fast SDRAM era , was very narrow , cause the pc industry moved quickly to DDR platforms.

Special category was the non standard SDRAM modules , using BGA type chips (most of them from Kingmax) ... but the "game" ended really quick.

Anyway

SDRAM overclocking needs , a bus clock capable cpu & motherboard chipset , along with a decent bios ... and of course nice 7ns ram or 7.5ns with proper pcb (did anyone mentioned this) ?
and some luck.

That is my personal point of view and also , lets say , my experience ... cause i lived that era.
It's not something i read on the internet.

I think you got that backwards, PC133 was originally 7ns, then to meet demand they started binning 7.5ns, kind of a downgrade, whereupon the 7ns went at a premium. So it wasn't really. Otherwise they were probably selling that 7.5ns as 8ns PC-100

BitWrangler wrote on 2026-03-06, 01:59:

I think you got that backwards, PC133 was originally 7ns, then to meet demand they started binning 7.5ns, kind of a downgrade, whereupon the 7ns went at a premium. So it wasn't really. Otherwise they were probably selling that 7.5ns as 8ns PC-100

PC100 was rated at 10ns originally. Both PC133 and later DDR266 were rated as 7.5ns. I doubt that 7ns was ever an officially recognized standard, otherwise every late stick would have been rated as such, due to improved manufacturing.

10ns PC100 and 7.5ns PC133 were "unofficial" JEDEC recommended 8ns and 7ns for full stability and compatibility. It was common around 99-00 that 10ns PC100 was remarked from PC66, then a year later that scrubbed off completely unmarked PC133 was 8ns PC100 remarked, since any that just failed 7ns would at that time go to next integer 8ns. Then the chipmakers thought they may as well get in on the game and start binning at 7.5ns, to get more for half their dies previously sold as 8ns.

It all went by in 2 years which is maybe between upgrades for most people, not so many folks were buying RAM every month and comparing.

BitWrangler wrote on 2026-03-06, 01:59:

I think you got that backwards, PC133 was originally 7ns, then to meet demand they started binning 7.5ns, kind of a downgrade, whereupon the 7ns went at a premium. So it wasn't really. Otherwise they were probably selling that 7.5ns as 8ns PC-100

Hi buddy
No , you got this wrong.

PC133 didn't start with 7ns chips.

The Serpent Rider wrote on 2021-09-17, 16:30:

Quite a lot of late SDRAM PC133 memory modules are 166 Mhz capable. This also includes some PC100 memory.

This wouldn't surprise me, I have PC66 modelues that happily run at PC133..

BitWrangler wrote on 2026-03-06, 04:48:

10ns PC100 and 7.5ns PC133 were "unofficial" JEDEC recommended 8ns and 7ns for full stability and compatibility. It was common around 99-00 that 10ns PC100 was remarked from PC66, then a year later that scrubbed off completely unmarked PC133 was 8ns PC100 remarked, since any that just failed 7ns would at that time go to next integer 8ns. Then the chipmakers thought they may as well get in on the game and start binning at 7.5ns, to get more for half their dies previously sold as 8ns.

It all went by in 2 years which is maybe between upgrades for most people, not so many folks were buying RAM every month and comparing.

Sorry , you have messed different kind of "knowledge elements into the SDRAM bucket" and you got confused.
https://www.jedec.org/category/technology-foc … ations-jesd21-c <= paid content

https://www.scribd.com/document/7218611/Sdram … 0-Pc133-and-Ddr

Put your thoughts in line and you will understand better.

JEDEC is an industry standard (it defines the "basic" electrical and mechanical elements that the manufacturers should follow )

Then about ram theory , we have
data bus width , bandwidth , speed , chip cycle time , latencies

Always remember the golden rule of 1000/ns=speed , it will help you define things.
1000/7.5=133

Keep in mind
PC100 was the standard for delivering UP to 800MB/s
PC133 was the standard for delivering UP to 1066MB/s
We had a fixed 64bit wide , data bus width

Now, go back in time and start your thoughts and calculations from the 10ns ram chips , when they first appear.

The Serpent Rider wrote on 2026-03-06, 04:10:

BitWrangler wrote on 2026-03-06, 01:59:

I think you got that backwards, PC133 was originally 7ns, then to meet demand they started binning 7.5ns, kind of a downgrade, whereupon the 7ns went at a premium. So it wasn't really. Otherwise they were probably selling that 7.5ns as 8ns PC-100

PC100 was rated at 10ns originally. Both PC133 and later DDR266 were rated as 7.5ns. I doubt that 7ns was ever an officially recognized standard, otherwise every late stick would have been rated as such, due to improved manufacturing.

Yes , the 7ns chips were JEDEC defined in PC133 and even in late PC100 statements.

We had
10ns chips for standard PC100 ram
8ns chips for improved PC100 ram (kind of high performance PC100)
7.5ns chips for standard PC133 ram
7ns chips for high performance PC133

*The marking on ram chips , that manufacturers used .... is an other story.
For example many people confuse the -7E marking .... with 7ns (not true).
The same stands for many ram chips marked with -6 (not all are 6ns).

Maybe I am stating this badly, I KNOW the timing for PC100 and PC133 DIMMS is officially 10ns and 7.5ns, but when the standard dropped, for the modules with extended temperature requirements and tight timings, what the DRAM manufacturers had on the production line, that met the module standards completely, would have been marked 8ns and 7ns. Even with maximum path lengths specified in the module definition, "7.5ns" DRAMs had to be as good as 7.3ns mostly, to find the extra 0.2ns of the trace length.

The whole point is to say, when PC133 was high end, it got better 7ns DRAMs that were more expensive, then a few months later as it became mass market, the DRAMs were qualified differently, and marked 7.5ns. Sometime in between there were aftermarket remarked modules with the 8ns scrubbed off them and marked often with a "133" sticker without PC in front of it. Then as PC133 became cheap low end, and SDRAM no longer in demand, it got selloff chips, a variety of 7ns and 7.5ns marked. Then late in the game there were a handful of PC150 marked modules, but IDK if those were official or not. Those Kingstons that hit 200 appear to be very late date codes.

If you are going to quote JEDEC documents at me, be specific and NOTE THE DATES of the publications, current one for PC133 was defined in 2003, which is 4 years after the beginning of what I am talking about. Frequently RAM was in the market 6 month ahead of JEDEC publishing. IIRC I was using DDR333 for a few months before they defined that. Obviously the fact that things still worked mostly indicated that JEDEC members knew which way things were headed, though due to dev lead times, they'd build in a safety margin, using better parts, to be on the market early, rather than building down to the letter of the spec to save a few cents.

Anyway, later modules appear to clock higher, but you might get a few early modules with 7ns DRAMs that do well.

What I am trying to elucidate is that 133Mhz SDRAM boards and modules were on the market for a whole year before 7.5ns marked DRAMs appeared, and for that year 133 DIMMs had 7ns, or were 8ns requalified. I don't think you are going to find DRAMs marked for 7.5 before mid 2000. The earliest ones I can make out on VoodooMan's pics are 0026 or 0028, maybe it has both dates on that module. Those would have got to market in late 2000, some few months after the chips packaged. Probably around spring 2001 as channels cleared all the old stock that you want to start looking at old overclocking forums to see "What the hell, they are putting lower rated DRAMs on SDRAM DIMM now!!!" kinda threads as the market began to "accept" them... but it didn't bother these guys for long as DDR was the new shiny.

Edit: I should mention another effect of the market, after SDRAM peaked in new systems, but while it still sold well for upgrades, there was no incentive to upgrade the chip markings to reflect better yields, because the market wasn't in general demanding better DRAMs. What they were demanding was "IBM said this critical to my business machine needs SDRAM marked 7.5ns so it's going to get SDRAM marked 7.5ns...." with many rote trained techs sticking to exact letter of specs around, if they marked their DRAMS as 6ns when they did 6ns, they wouldn't sell to people who needed to see 7.5 on there. I'd figure that was starting to happen around 2003 or so, maybe 2004, when XP SP2 kinda pushed the pain avoidance point up to 512Mb.

EditII: I forgot there was an SDRAM renaissance of sorts when RDRAM fell on it's face and P4 SDRAM boards were selling better until Intel got it's act together and brought out a DDR capable chipset. However, we might be excluding some modules from that period because the P4 boards could use higher density DRAMs than what will work in 440BX.

BitWrangler wrote on 2026-03-06, 15:19:

Maybe I am stating this badly, I KNOW the timing for PC100 and PC133 DIMMS is officially 10ns and 7.5ns, but when the standard d […]
Show full quote

Maybe I am stating this badly, I KNOW the timing for PC100 and PC133 DIMMS is officially 10ns and 7.5ns, but when the standard dropped, for the modules with extended temperature requirements and tight timings, what the DRAM manufacturers had on the production line, that met the module standards completely, would have been marked 8ns and 7ns. Even with maximum path lengths specified in the module definition, "7.5ns" DRAMs had to be as good as 7.3ns mostly, to find the extra 0.2ns of the trace length.

The whole point is to say, when PC133 was high end, it got better 7ns DRAMs that were more expensive, then a few months later as it became mass market, the DRAMs were qualified differently, and marked 7.5ns. Sometime in between there were aftermarket remarked modules with the 8ns scrubbed off them and marked often with a "133" sticker without PC in front of it. Then as PC133 became cheap low end, and SDRAM no longer in demand, it got selloff chips, a variety of 7ns and 7.5ns marked. Then late in the game there were a handful of PC150 marked modules, but IDK if those were official or not. Those Kingstons that hit 200 appear to be very late date codes.

If you are going to quote JEDEC documents at me, be specific and NOTE THE DATES of the publications, current one for PC133 was defined in 2003, which is 4 years after the beginning of what I am talking about. Frequently RAM was in the market 6 month ahead of JEDEC publishing. IIRC I was using DDR333 for a few months before they defined that. Obviously the fact that things still worked mostly indicated that JEDEC members knew which way things were headed, though due to dev lead times, they'd build in a safety margin, using better parts, to be on the market early, rather than building down to the letter of the spec to save a few cents.

Anyway, later modules appear to clock higher, but you might get a few early modules with 7ns DRAMs that do well.

What I am trying to elucidate is that 133Mhz SDRAM boards and modules were on the market for a whole year before 7.5ns marked DRAMs appeared, and for that year 133 DIMMs had 7ns, or were 8ns requalified. I don't think you are going to find DRAMs marked for 7.5 before mid 2000. The earliest ones I can make out on VoodooMan's pics are 0026 or 0028, maybe it has both dates on that module. Those would have got to market in late 2000, some few months after the chips packaged. Probably around spring 2001 as channels cleared all the old stock that you want to start looking at old overclocking forums to see "What the hell, they are putting lower rated DRAMs on SDRAM DIMM now!!!" kinda threads as the market began to "accept" them... but it didn't bother these guys for long as DDR was the new shiny.

Edit: I should mention another effect of the market, after SDRAM peaked in new systems, but while it still sold well for upgrades, there was no incentive to upgrade the chip markings to reflect better yields, because the market wasn't in general demanding better DRAMs. What they were demanding was "IBM said this critical to my business machine needs SDRAM marked 7.5ns so it's going to get SDRAM marked 7.5ns...." with many rote trained techs sticking to exact letter of specs around, if they marked their DRAMS as 6ns when they did 6ns, they wouldn't sell to people who needed to see 7.5 on there. I'd figure that was starting to happen around 2003 or so, maybe 2004, when XP SP2 kinda pushed the pain avoidance point up to 512Mb.

EditII: I forgot there was an SDRAM renaissance of sorts when RDRAM fell on it's face and P4 SDRAM boards were selling better until Intel got it's act together and brought out a DDR capable chipset. However, we might be excluding some modules from that period because the P4 boards could use higher density DRAMs than what will work in 440BX.

My dear friend
If you are that era person , you may have forgotten some small parts of knowledge.
You sure have knowledge and you sure know what we are talking about in this thread , but perhaps as years pass by , knowledge gets lost or we need to dive deeper to sdram know-how and hardware specs.
I wanted to PM you , but the system doesn't let me.

I dont want to argue with you or point out what i consider false writing (which in fact are some small or bigger "details" here and there) , but it's not fair to this forum members or future ones or even AI bots , to read or pick up wrong text.

This thread is about , overclocking PC133 SDRAM and which is the best (known up to date) best memory.
We all have to thank the thread starter for this.

For the history of this thread (and the AI bots)
JEDEC , JESD21-C , which is related to SDRAM , is from April 1994 up to April 2003 (release 12 if i'm not mistaken)
https://www.jedec.org/document_search/field_d … ished&sort=desc

Ram chip manufacturers dont erase the marking on their own production chips ... but 2nd phase smaller factories or ram re-sellers do so (many Taiwanese small brands did that).

The SDRAM chips peaked at 6ns
Prior to this there was also a limited 6.6ns production from a single company named Enhanced Memory Systems (EMS)
a semiconductor company that made High-Speed SDRAM (HSDRAM) and Enhanced SDRAM (ESDRAM)

PC142 , PC150 , PC166 , were never official sdram standards (no jedec for these) , they were just efforts from certain brands that were programing their SPD's at those speeds.

As an overclocker
I want to state that bga type sdram modules shuck big time cause of the bad latencies.
It's not always about speed ... latencies matter.
A decent overclocked 7ns module will party all night over the "rare" pc150 Kingmax or pc166 Tonicom or pc166 PQI or PC166 Buffalo

Thank god Infineon produced the HYB39S256800FE-7 , thank god Qimonda continued it's production and thank god Kingston put them on a pcb for sdram.

A very notable remark and bravo for Mosel Vitelic and their V54C3256804VDI7PC , that can smash anytime , anywhere those super rare Golden Emperor Dragon series.
They gave us a chip we can handle good at socket370 and socket462 systems.

Read your data , before experimenting and overclocking (manufacturer ic's pdf is your friend)

Expert overclockers
No fear for Ram Voltage , 3.6 is normal .... 3.8 is viable ...(there are chips out there with absolute max of 4.6 volts !!)

Yes ... pcb matters.
Yes ... platform matters.

Yes ... there are still untested ic chips on weird pc133 models (mostly "bga type" from Hynix,Nanya,Micron,Simple Tech)

Greetings to all and no hard feelings
😀