Patriot Hellfire M.2 240GB NVMe SSD Review

[Guest]

The Patriot Hellfire M.2 has new performance-enhancing firmware and a new price point. Today we take our first look at the 240GB model and update the performance of the 480GB with the latest firmware release to see how fast Hellfire really is.

Patriot Hellfire M.2 240GB NVMe SSD Review : Read more

 

[Game256]

"The Hellfire M.2 is currently the only NVMe SSD to target the middle, but that will change next month when Samsung ships the 960 EVO to resellers"

The release date of Samsung 960 EVO/PRO is October. Or you know "something"?

 

[Game256]

Another meaningless and overpriced drive. The only thing that may help all these vendors to compete with Samsung is the price but usually they constantly overstate it, making it too close to Samsung 950 Pro. Samsung is going to kill them all with 960 EVO.

 

[Game256]

Btw some Polish and German tech sites already received samples of Samsung 960 Pro so it's interesting that you didn't received them yet.

https://scontent-waw1-1.xx.fbcdn.net/v/t1.0-0/s480x480/14729397_2937997012914136_7324007564864990086_n.jpg?oh=e0996b629cf74e7be362bd42809d54f2&oe=58AA4E34

https://scontent-waw1-1.xx.fbcdn.net/t31.0-8/14612499_10154535560642207_9118950000337195147_o.jpg

 

[CRamseyer]

As of an hour ago I can no longer talk about the new Samsung SSDs.

 

[Brian_R170]

"The Patriot Hellfire M.2 and Intel 600p are the first products to really burst the NVMe pricing bubble." Are you speculating on future pricing? The statement is hardly true based on actual retail or suggested retail pricing. I checked the retail pricing of currently-available M.2 NVMe SSDs and all but 2 cluster together with pricing differences of less than 10%. The outliers are the Intel 600p and the Plextor M8Pe.

 

[shrapnel_indie]

I guess bursting the bubble depends on the source of your part: At the beginning of the article, the viewed pricing on my end was more in competition to Samsung's M.2 PCIe offerings a few months ago. At the end of the article, it was more in-line for the mid-range to high-range SATA drives of comparable size.

 

[JustAnITGuy]

PLEASE ALWAYS include a standard 7200RPM 3.5" drive and a 7200RPM 2.5" drive with ANY SSD benchmarks. The vast majority of people still have these old drives, and if this is to educate people to purchase products, It would be nice to see a side-by-side comparison and not have to go to separate benchmark articles or websites.

 

[CRamseyer]

The results are so much different we can't really put them in the same chart together. Maybe in the future we will publish an article with a mainstream HDD, SSD and NVMe SSD for a comparison.

 

[steve4king]

"It turns out the Patriot Hellfire M.2 is not the product we originally thought it was. After some tuning, the application performance is up and this drive is ready to take on the emerging mainstream NVMe market."

What do you mean? The product you originally thought it was was based on the original firmware.. right? The 2.1 firmware is about the same, or noticeably worse than the 2.0 firmware in every bench..

 

[steve4king]

Under sequential read performance, I think you have a typo:
"The 240GB outperforms the Samsung PM961, and Intel's 600p nips at the heels of the premium Samsung 950 Pro 256GB."

I think you meant to say:
The 240GB outperforms the Samsung PM961 and Intel's 600p, and nips at the heels of the premium Samsung 950 Pro 256GB.

 

[CRamseyer]

I'm not too concerned with the synthetic tests. They are a guide that gives us insight into real-world performance. The real-world tests give us a better understanding on what to expect with the drive in your system as an operating system device. The difference is more data on the device, increased background activity and latency with mixed workloads.

A company can tune a drive for excellent synthetic results but perform poor in applications. There are a number of examples but many of the companies have moved away from tuning for reviews after we started showing mixed workload results. FW 2.1 seems to have increased the amount of CPU power dedicated to background activity and that is why the application performance is higher.

This is the type of performance we expected from the Hellfire the first time but the early firmware might have been tuned more for benchmarks. Without the firmware team to verify we can't say for sure but we can say the new firmware will deliver higher performance with the drive in your system as a boot device.

 

[RedJaron]

The release date of Samsung 960 EVO/PRO is October. Or you know "something"?

Btw some Polish and German tech sites already received samples of Samsung 960 Pro so it's interesting that you didn't received them yet.

Are you speculating on future pricing? The statement is hardly true based on actual retail or suggested retail pricing. I checked the retail pricing of currently-available M.2 NVMe SSDs and all but 2 cluster together with pricing differences of less than 10%. The outliers are the Intel 600p and the Plextor M8Pe.

You need to understand that reviews aren't published the same day the writing is completed. They're put in a queue where the editors decide which ones will be put up on which dates. It's not uncommon for a review to sit in the queue for a few weeks in certain situations.

 

[popatim]

I would love to see an SATA SSD added to the real world applications benchmark to see how much an improvement an nvme drive really is.

What are we measuring those service times in Anyways? Milliseconds, Micro???

 

[Co BIY]

Agree with JustanITGuy. With the application performance difference of all these drives being less than 2% it makes sense to chose a class of performance and then purchase on price (or longevity, or power consumption).

We need that class comparison article !

 

[zodiacfml]

Thanks for always including the Real World performance. It seems that these M.2 drives performs similarly. I just can't shake that Samsung drives has better endurance though.

 

[HERETIC-1]

Hey Chris,Think you nailed it with-
"This is a road we've traveled before; we will call it the Phison Highway. The Patriot Hellfire will receive updates on the way to an undisclosed destination. History tells us no one knows where we may stop along the way, or how finished the E7 will ever be. The S10 never delivered on all of its promises, and it has been on the road for more than two years."

My take was OCZ/Phison have been using us as "crash test dummies" with S10.

 

[babernet_1]

Why do you never run boot time tests?

How about game loading times?

That is what we are interested in.

 

[Jeff_106]

It's really hard to see how a drive that offers only marginal performance gains that may not be seen by the average user for 100% price premium (with low end SSDs going for under $60) does not have much appeal. The performance gains on going from a spinning drive to an SSD are revolutionary while the gains going from SATA to NVME are only evolutionary and then will only appeal at the leading edge (like most readers of this site.

 

[nebun]

interesting little fast drive...a little overpriced, but fast nevertheless

 

[RedJaron]

Why do you never run boot time tests?

How about game loading times?

That is what we are interested in.

Because once you move to flash storage, boot time is more a factor of your mboard and OS. How long does your board take to POST? How long did you delay it to recognize F2 / Del to get into the BIOS? Does it support quick boot and have you enabled it? And while a bare-bones Windows installation can fully load in a few seconds, most people have all sorts of background apps and services that take extra time to fully load and run, much of which starts being a function of CPU and RAM as well as SSD. Don't forget the time to enter your login credentials.

So while I can setup a computer to boot in under 10 seconds, real world application means it's closer to 20 seconds for most people. The difference in flash speed would result in less than a second difference, meaning under a 5% performance difference.

Game loading will see the same thing. Most have splash screens or intro videos, many which can't be skipped. Even if they can, it requires user intervention, which is basically impossible to get consistent. Once in the game and loading a save, a lot of games also have mandatory load screens with fixed durations. Once again, the difference of flash drives in how fast games load is almost negligible in comparison with the rest of the factors.

 

[babernet_1]

Why do you never run boot time tests?

How about game loading times?

That is what we are interested in.

Because once you move to flash storage, boot time is more a factor of your mboard and OS. How long does your board take to POST? How long did you delay it to recognize F2 / Del to get into the BIOS? Does it support quick boot and have you enabled it? And while a bare-bones Windows installation can fully load in a few seconds, most people have all sorts of background apps and services that take extra time to fully load and run, much of which starts being a function of CPU and RAM as well as SSD. Don't forget the time to enter your login credentials.

So while I can setup a computer to boot in under 10 seconds, real world application means it's closer to 20 seconds for most people. The difference in flash speed would result in less than a second difference, meaning under a 5% performance difference.

Game loading will see the same thing. Most have splash screens or intro videos, many which can't be skipped. Even if they can, it requires user intervention, which is basically impossible to get consistent. Once in the game and loading a save, a lot of games also have mandatory load screens with fixed durations. Once again, the difference of flash drives in how fast games load is almost negligible in comparison with the rest of the factors.

I watched a youtube (Yes, I know) video where a guy got several seconds quicker with a NVMe vs SATA SSD. I think it would be very worthwhile to set up some sort of standardized test with a clean load and, maybe, a half dozen to dozen normal boot type applications. You can set up for no password entry and a standard, say, one-second delay for BIOS. I would think it would be more than a fraction of a second difference for a NVMe.

Then you could find several games that are directly slowed down by the SSD and use those as a reference.

What you are saying is almost like saying "Hey, Minsweeper runs at 100 FPS with Intel's integrated graphics, so why test separate graphics cards?"

 

[CRamseyer]

Here is the problem with boot to Windows style tests. It boot time really depends on how many files (DLLs mostly) you have coming up with Windows. We all know that Windows boot times get longer with the OS install age. That is the result of users loading more software on the drives. If I were to test with a fresh install of Windows then the results would need to be accurate to thousandths of a second. There isn't a tool accurate enough to show a meaningful difference. If I were to show an old installation of Windows then it would be unrealistic for users standing up a new installation. It really comes down to not having a tool accurate enough to show split second differences. Even if a tool existed though, I don't think the difference between one product and another would be enough to sway your buying decision.

We do run two game loading sequences in the PCMark 8 section. World or Warcraft and Battlefield 3. I proposed and pushed for BF3 to be in the storage test suite with Futuremark. My little contribution to storage history. I didn't have the foresight to ask for more detailed results. The test only goes out to tenths of a second.

 

[RedJaron]

What you are saying is almost like saying "Hey, Minsweeper runs at 100 FPS with Intel's integrated graphics, so why test separate graphics cards?"

Except it's really not. I'm not saying Minesweeper is a close representative of the majority of games nor am I saying the Intel iGPU is a good analogue for most GPUs. I'm saying from a perceived user experience perspective, most flash drives perform close to each other.

Yes, some controllers and flash quality can theoretically perform much faster than others when all other restrictions are removed from the equation. So when you're using synthetic testing for IOPS and massive sequential read speed, you see a big gap. Those performance benefits aren't noticed near as much in real world unless you're moving mass data around. Basic functions like machine start up and application launch won't appear noticeably different. Look at the application specific benches Chris posted in the review. They're nearly flat.

If you still don't believe me, or don't believe Chris in how fast boot bloat happens, I tested my X99 testbed tonight with both a Samsung 840 Pro ( AHCI SATA ) and Toshiba RDV400 ( NVMe 3.0 x4 ). Using identical images I timed them from the boot drive select menu, thus removing POST and other delays. Best tool I had was a stopwatch on my phone. Both booted to Win10 desktop in 8 seconds, +/- 100 ms for my reflex response time. Is it possible the NVMe is 200 ms faster in booting? Sure. But that's only 2.5% improvement at best, not to mention it's within margin of error so it's largely irrelevant. You'd need a boot time of 40 seconds to get even a 1 second difference, which again wouldn't be noticeable.

 

[babernet_1]

What you are saying is almost like saying "Hey, Minsweeper runs at 100 FPS with Intel's integrated graphics, so why test separate graphics cards?"

Except it's really not. I'm not saying Minesweeper is a close representative of the majority of games nor am I saying the Intel iGPU is a good analogue for most GPUs. I'm saying from a perceived user experience perspective, most flash drives perform close to each other.

Yes, some controllers and flash quality can theoretically perform much faster than others when all other restrictions are removed from the equation. So when you're using synthetic testing for IOPS and massive sequential read speed, you see a big gap. Those performance benefits aren't noticed near as much in real world unless you're moving mass data around. Basic functions like machine start up and application launch won't appear noticeably different. Look at the application specific benches Chris posted in the review. They're nearly flat.

If you still don't believe me, or don't believe Chris in how fast boot bloat happens, I tested my X99 testbed tonight with both a Samsung 840 Pro ( AHCI SATA ) and Toshiba RDV400 ( NVMe 3.0 x4 ). Using identical images I timed them from the boot drive select menu, thus removing POST and other delays. Best tool I had was a stopwatch on my phone. Both booted to Win10 desktop in 8 seconds, +/- 100 ms for my reflex response time. Is it possible the NVMe is 200 ms faster in booting? Sure. But that's only 2.5% improvement at best, not to mention it's within margin of error so it's largely irrelevant. You'd need a boot time of 40 seconds to get even a 1 second difference, which again wouldn't be noticeable.

Hmm, interesting data. My son is ready to upgrade from his 256G Samsung SSD. We were going to go with a 500GB 960 EVO. Maybe we should stick to a SATA M.2 to save a lot of money. Maybe get a 1TB SATA M.2.

Thanks.