Tuesday, September 15th 2026

Goodram Introduces Master PCIe 5.0 SSD Series With Speeds Up to 14,000 MB/s

Goodram is introducing Master, a consumer NVMe PCIe 5.0 SSD designed for next-generation platforms. The new model is aimed at gamers, content creators, professionals and PC enthusiasts who demand top performance when working with large files, the latest games and AI-powered applications.

Thanks to the PCIe 5.0 x4 interface, Goodram Master delivers sequential read speeds of up to 14,000 MB/s, write speeds of up to 13,000 MB/s and performance of up to 2,000,000 IOPS. These specifications translate into shorter application and game loading times, faster transfers of large files and high system responsiveness even during intensive multitasking.
The new SSD has been designed for modern PCs equipped with the latest processors and motherboards supporting the PCIe 5.0 standard. The high bandwidth of the interface helps users take better advantage of the capabilities of today's platforms while providing additional performance headroom for the years ahead.

Goodram Master is also well suited to workloads involving local artificial intelligence. Fast access to AI models, datasets and working files helps reduce environment setup times and, when combined with a powerful GPU and sufficient VRAM, enables more efficient use of local AI tools.

The SSD supports Microsoft DirectStorage, which can reduce loading times and accelerate texture streaming in compatible games.

Another key feature of the new model is its energy efficiency. According to Goodram's internal testing, Master consumes up to 27% less power than selected previous-generation PCIe 5.0 SSDs offering a comparable level of performance. Its single-sided design has been developed to work with heatsinks integrated into modern motherboards, helping to dissipate heat effectively during sustained workloads.

Goodram Master will be available in 1 TB and 2 TB capacities. The 1 TB version offers sequential read speeds of up to 14,000 MB/s and write speeds of up to 9,000 MB/s, while the 2 TB model delivers up to 14,000 MB/s read and 13,000 MB/s write speeds, as well as random read performance of up to 2,000,000 IOPS. Both versions are covered by a 5-year limited warranty.

Key specifications of Goodram Master
  • PCIe NVMe 5.0 x4
  • sequential read speeds of up to 14,000 MB/s
  • sequential write speeds of up to 13,000 MB/s (2 TB)
  • performance of up to 2,000,000 IOPS
  • Microsoft DirectStorage support
  • Host Memory Buffer (HMB) technology
  • up to 27% lower power consumption
  • single-sided M.2 2280 design
  • capacities: 1 TB and 2 TB
  • 5-year limited warranty subject to the TBW limit
Source: Goodram
Add your own comment

16 Comments on Goodram Introduces Master PCIe 5.0 SSD Series With Speeds Up to 14,000 MB/s

#1
shuffy_losys
Aside from flash memory supply still being a problem...

These improvements are at best benchmark values in most consumer environments. I'm pretty sure we've hit a point where most storage-to-RAM transfer times are a few seconds at most and the rest of the "loading" is all the processing that needs to happen to get the thing up and running.
Posted on Reply
#2
Evrsr
shuffy_losysAside from flash memory supply still being a problem...

These improvements are at best benchmark values in most consumer environments. I'm pretty sure we've hit a point where most storage-to-RAM transfer times are a few seconds at most and the rest of the "loading" is all the processing that needs to happen to get the thing up and running.
It is worse than that. I have a new PCIe 4.0 drive from WD and the damn thing sits at 30-40MB/s writes if you fill the cache. Download 15GB and watch the other 15GB slow to a crawl. An old 3.0 Lexar drive I have manages at least 70+MB/s and the SN770 I think was around 500MB/s.

Most of my old SATA drives were faster than this WD drives that are put on laptops. If the QLC drives can do the cache management in an aggressive way, they are ok.

At least the random reads seem to have doubled on most drives, so that is a 2x boost to OS workloads and stuff like that.
Posted on Reply
#3
vgm
EvrsrIt is worse than that. I have a new PCIe 4.0 drive from WD and the damn thing sits at 30-40MB/s writes if you fill the cache. Download 15GB and watch the other 15GB slow to a crawl. An old 3.0 Lexar drive I have manages at least 70+MB/s and the SN770 I think was around 500MB/s.

Most of my old SATA drives were faster than this WD drives that are put on laptops. If the QLC drives can do the cache management in an aggressive way, they are ok.

At least the random reads seem to have doubled on most drives, so that is a 2x boost to OS workloads and stuff like that.
Disable Write cache on the drive to fix the issue in sustained writes with steam or large files. Its a fault in firmware that handles HMB which doesn't get cleared or updates until you forcibly clear standby memory using RamMap or ISLC. WD refuses to acknowledge it since SN570 years ago.
Posted on Reply
#4
Let3r
Nomad76Goodram Master is also well suited to workloads involving local artificial intelligence. Fast access to AI models, datasets and working files helps reduce environment setup times and, when combined with a powerful GPU and sufficient VRAM, enables more efficient use of local AI tools.
Of course we get to this. 😞
Posted on Reply
#5
bonehead123
Ram, bam, no thank ya ma'am :D

I'll stick with my Gen 4 850x's for now and will be happier for it, since I have a large stash of them.....
Posted on Reply
#6
Evrsr
vgmDisable Write cache on the drive to fix the issue in sustained writes with steam or large files. Its a fault in firmware that handles HMB which doesn't get cleared or updates until you forcibly clear standby memory using RamMap or ISLC. WD refuses to acknowledge it since SN570 years ago.
Do you have details on this? Cause depending on the type of cache you say, it can cause issues with data integrity.

If you are suggesting to disable Write Caching, that is very dangerous because you will loose the data if there is a crash. You should never do that unless you are using things like RAID arrays with batteries.
Posted on Reply
#7
Wasteland
EvrsrIt is worse than that. I have a new PCIe 4.0 drive from WD and the damn thing sits at 30-40MB/s writes if you fill the cache. Download 15GB and watch the other 15GB slow to a crawl. An old 3.0 Lexar drive I have manages at least 70+MB/s and the SN770 I think was around 500MB/s.

Most of my old SATA drives were faster than this WD drives that are put on laptops. If the QLC drives can do the cache management in an aggressive way, they are ok.

At least the random reads seem to have doubled on most drives, so that is a 2x boost to OS workloads and stuff like that.
What model is the WD drive with the problem? SN770 is a WD drive, for example, lol.
Posted on Reply
#8
Evrsr
WastelandWhat model is the WD drive with the problem? SN770 is a WD drive, for example, lol.
WD SN5000S, seems Acer laptops ship with these now. Worse NVMe drive I had despite being market as PCIe 4.0 drive.
Posted on Reply
#9
Wasteland
EvrsrWD SN5000S, seems Acer laptops ship with these now. Worse NVMe drive I had despite being market as PCIe 4.0 drive.
Ah yeah. QLC job. Would not touch with a ten foot pole.
Posted on Reply
#10
Evrsr
WastelandAh yeah. QLC job. Would not touch with a ten foot pole.
That and page size of those chips seems to already be at 16KiB when NTFS MFT writes are 1-4KiB, with no really good way to match what is underneath better (AFAIK). Makes the problem much worse due to 4-16x write amplification for metadata.

MS really needs to start improving what they are doing with filesystems these days. Some Linux FS can already pack the data when writing the metadata/inode, improving performance and reducing writes.
Posted on Reply
#11
vgm
EvrsrDo you have details on this? Cause depending on the type of cache you say, it can cause issues with data integrity.

If you are suggesting to disable Write Caching, that is very dangerous because you will loose the data if there is a crash. You should never do that unless you are using things like RAID arrays with batteries.
I think its actual the reverse is true. Write caching is usually disabled to write directly to media which is quite slower if there's COW (Copy on Write) or Allocate-on-Flush (AoF). On linux write caching was disabled by default (until a decade ago) as it was buggy and caused some hdd (seagate barracuda/hitachi 200GB-500GB) as Xubuntu 11/12 corrupted my disk on huge file copy and write from internal to external HDD losing 10-12GB of data.
Windows doesn't flush/refreshes HMB and standby frequently to be very conservative. The same drive on Linux Bazzite/Fedora/Nobara/Devuan/Debian Flies and I don't even see a loading screen or progress window.
I think the second option under Write Cache that is, disable Windows Write Cache Buffer flushing is the cause for data integrity if you don't have backup power/UPS. On laptop its unaffected.
Posted on Reply
#12
Evrsr
vgmI think its actual the reverse is true. Write caching is usually disabled to write directly to media which is quite slower if there's COW (Copy on Write) or Allocate-on-Flush (AoF). On linux write caching was disabled by default (until a decade ago) as it was buggy and caused some hdd (seagate barracuda/hitachi 200GB-500GB) as Xubuntu 11/12 corrupted my disk on huge file copy and write from internal to external HDD losing 10-12GB of data.
Windows doesn't flush/refreshes HMB and standby frequently to be very conservative. The same drive on Linux Bazzite/Fedora/Nobara/Devuan/Debian Flies and I don't even see a loading screen or progress window.
I think the second option under Write Cache that is, disable Windows Write Cache Buffer flushing is the cause for data integrity if you don't have backup power/UPS. On laptop its unaffected.
Yes I didn't understand correctly, you are suggesting disabling the flushing.

I will try what you say and disable caching, which should always be fine and fast. It means changes only are committed to metadata (on modern FSs) after the drive reports cache data fully written out to disk.
So on a crash you either get old data or new one, it is hard to get corruption but can still happen.
Posted on Reply
#13
vgm
EvrsrYes I didn't understand correctly, you are suggesting disabling the flushing.

I will try what you say and disable caching, which should always be fine and fast. It means changes only are committed to metadata (on modern FSs) after the drive reports cache data fully written out to disk.
So on a crash you either get old data or new one, it is hard to get corruption but can still happen.
Disabling write cache also disable windows write cache flushing as the changes are committed to the media directly and nvme is fast. Data loss is virtually zero unless there's brownouts or overload.
Unless you're affected by slow write on specific WD/Sandisk NVMe drives under heavy writes disabling you won't notice much difference.
Posted on Reply
#14
Evrsr
vgmDisabling write cache also disable windows write cache flushing as the changes are committed to the media directly and nvme is fast. Data loss is virtually zero unless there's brownouts or overload.
Unless you're affected by slow write on specific WD/Sandisk NVMe drives under heavy writes disabling you won't notice much difference.
But the fact it could be the same speed is weird. Cause with HMB it can write bigger blocks. I assume with no kind of cache, you are increasing your write amplification by a lot.

Maybe the controller can do some form of management on internal cache that is enough to work ok.

Will try it out on my laptop and see how it goes.
Posted on Reply
#15
vgm
EvrsrBut the fact it could be the same speed is weird. Cause with HMB it can write bigger blocks. I assume with no kind of cache, you are increasing your write amplification by a lot.

Maybe the controller can do some form of management on internal cache that is enough to work ok.

Will try it out on my laptop and see how it goes.
Yes, it will be almost the same as incoming writes won’t be soaked by memory and there’s less controller work.
HMB under heavy 4k writes simply produces worst case scenario for QLC and higher.
Posted on Reply
#16
Evrsr
vgmYes, it will be almost the same as incoming writes won’t be soaked by memory and there’s less controller work.
HMB under heavy 4k writes simply produces worst case scenario for QLC and higher.
Actually tried some stuff today. This particular SN5000S sometimes goes as low as 15MB/s for several seconds at a time. Maybe it is overheating.

Tried your suggestions but did not change everything. What I found is Win11 disabling cache-flushing by default... (or by OEM configuration). W10 definitely does not do that.

Looked for HBM registry settings and it has none defined, so maybe it is just really bad QLC with an overheating controller. (85ºC)
Posted on Reply
Add your own comment
Nov 17th, 2026 12:34 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts