• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Performance issue with new intel 660p 1tb

So you guys are implying i should REMOVE the heatsink from my HP EX950 2TB and ADATA sx8200 pro 1TB?
(they are both sitting under the massive heatsink of an asus hyper m2 daughter card, though the fan is NOT running)
 
So you guys are implying i should REMOVE the heatsink from my HP EX950 2TB and ADATA sx8200 pro 1TB?
(they are both sitting under the massive heatsink of an asus hyper m2 daughter card, though the fan is NOT running)
IMO No, if it runs cool leave it alone. The controller is meant to be kept cool but the NAND should be allowed to heat up. Either way probably doesn't matter. These things are also meant to be installed in ultrabooks and they get really hot in that environment.
 
terrible performance!! Return it while you still can and buy something better, there are plenty of brands with much better performance....QLC is just a no go!

But there isn't anything with this performance at this price point and size. The 660p performs very well in reads, but writes hurt it. It's just fine in most real world use.
 
But there isn't anything with this performance at this price point and size. The 660p performs very well in reads, but writes hurt it. It's just fine in most real world use.
Yes there is. The E12 based drives.
Sabrent Rocket, Corsair MP510, PNY 3030, Silicon Power 34A80, Inland 1TB X4, Seagate FireCuda, Team Group MP34, MyDigitalSSD BPX Pro
 
Tried my video folder from my intel 660 to my usb 3 2.5" harddrive (backup), it peaker just around 100 mbs, mabye an avarage of 85-90 mb/s it was never more than 41 c hot
 
what you tried would never heat it up as the external drive is bottlenecking the speed.
 
Yes there is. The E12 based drives.
Sabrent Rocket, Corsair MP510, PNY 3030, Silicon Power 34A80, Inland 1TB X4, Seagate FireCuda, Team Group MP34, MyDigitalSSD BPX Pro

Not a single one of those drive is at the price point of the 660p. The 660p is still the cheapest 1TB NVMe SSD on the market. The closest is the Crucial P1, and it suffers from the same performance issues since it is a QLC drive as well.

Tried my video folder from my intel 660 to my usb 3 2.5" harddrive (backup), it peaker just around 100 mbs, mabye an avarage of 85-90 mb/s it was never more than 41 c hot

NVMe drives tend to get hotter during writes, not reads.
 
Not a single one of those drive is at the price point of the 660p. The 660p is still the cheapest 1TB NVMe SSD on the market. The closest is the Crucial P1, and it suffers from the same performance issues since it is a QLC drive as well.
660P 1TB is about $100. I have routinely seen the Phison at $105.
 
Check and make sure that windows has trimmed it open defrag and it'll tell you when it was last trimmed

even though it's set to trim it everyday on my machine when I checked it last it hadn't been trimmed at all for 17 bloody days
 
660P 1TB is about $100. I have routinely seen the Phison at $105.

The 660p is usually closer to $90 than $100, and I haven't been able to find a Phison drive under $110.
 
660P 1TB is about $100. I have routinely seen the Phison at $105.

In denmark where i live the intel 660p 1tb is by far the cheapest 1 tb nvme

I could just by a 1tb sata m.2 ssd and call it a day but i want to buy something faster or else it was a waste of money selling my sata ssd and just buy a sata m.2 ssd


Samsung 860 EVO M.2 2280 SSD - 1TB cost the same as a A-Data XPG SX8200 Pro M.2 SSD - 1TB it's insane not to buy a nvme ssd (x2 or x4 i don't care as long as it has stable performance and are faster than a sata ssd)

If i could get a 2 tb samsung 860 evo m.2 for the same price as a samsung 970 evo 512 it would be the only situation where i would consider buying a sata m.2 ssd over a nvme ssd, that is not gonna happen, prices are not going to be that low.

Check and make sure that windows has trimmed it open defrag and it'll tell you when it was last trimmed

even though it's set to trim it everyday on my machine when I checked it last it hadn't been trimmed at all for 17 bloody days

Trimmed? Defrag only tells me if the program has optimized my ssd and using the defrag and optimizing program is something you absolutely shouldn't do

It effects the lifespan of the ssd and won't make it faster
 
So what to buy instead?

The samsung 970 evo 500gb i had a few month ago was sooooooo hot, close to 70c when i had it under my heatsink (no sticker removed), couldn't live with this kind of high temps close to my gpu.
I've got the same drive. Sure it gets toasty under load... mine will go as high as 75c with no heatsink. Briefly.

But it's not even a thing.... as soon as activity drops back down it's at 40c. Near instantly.

No gpu temp problems either. Haven't compared but I'm at 55-65 lately, depending on the title. Done so many builds with them adjacent to GPUs with no problems. I mean... when the GPU is working its hardest, ie gaming, the SSD isn't exactly getting nailed with massive r/w operations. Little dings here and there.

But even if they were both loaded down together it wouldn't matter much.

I mean... it's not just temp. It's mass too. How much heat capacity could a piece of silicon the size of a stick of gum possibly have? Point is... yes the temperature gets high. But that doesn't mean it's a lot of actual heat energy. Something so small can only gather so much... even though temp rises faster. And in reality, that happens simply because it is smaller... it doesn't take as much heat energy to raise the temperature of such a small object.

Which will do more damage? A penny heated to 200c? Or a 4' piece of rebar heated to 200c? Which carries more heat?
 
Last edited:
terrible performance!! Return it while you still can and buy something better, there are plenty of brands with much better performance....QLC is just a no go!

Well that is why I have I have 4 of these in RAID 0. The fact of the matter is that almost all NVME drives will do that once they have reached a certain capacity. It is the basic difference between the 970 Pro and EVO.
 
Toms Hardware
And two sentences below your passage it said this..........

"Given the size of the 660p's SLC caches, you will likely never encounter this condition."

ATTO or any of these programs did not write 130GB worth of data. That doesn't seem to be the issue.
 
And two sentences below your passage it said this..........

"Given the size of the 660p's SLC caches, you will likely never encounter this condition."

ATTO or any of these programs did not write 130GB worth of data. That doesn't seem to be the issue.

Understood, the drive is set at 140GB written before the Cache is filled. I have noticed too that I am not getting the full write speeds on my SX8200 Pro RAID 0 array I think when I ran a benchmark the read speeds were in the 1600 range but rights were just short of 6000. I will look into it when I get home and post the results. I do have to sell some parts today though.
 
The 660p is usually closer to $90 than $100, and I haven't been able to find a Phison drive under $110.
Well I found one $108, https://www.microcenter.com/product...80-pcie-nvme-30-x4-internal-solid-state-drive

... But you had said that it was inferior, despite it being the exact same physical hardware (PCB, controller, NAND, Ram) and firmware as every other Phison reference drive.
Your comment was here:
I think most of their stuff is just rebranded cheap chinese stuff that you can get for slightly cheaper directly off aliexpress. I don't think they actually manufacture anything they sell. And that's why the quality can be hit or miss. Sometimes they get a good product from one manufacturer, but another product they are getting from a completely different manufacturer is bad.

And I bet these SSDs are the same deal. Random SSDs from China that they put their sticker on. It's likely using chips that failed QA at one of the brand name manufacturers, but is good enough for an off-brand.


E12 Reference review here: https://www.anandtech.com/show/13080/the-phison-e12-nvme-ssd-controller-preview
If you want a mainstream branded drive then get a HP EX 920 $115/1TB or EX 950 $130/1TB those are also in the same ball park price range. Those are similar in speed to 970, the Phison and the Adata XPG 8200 Pro.

Another thing is the 660P's QLC endurance is terrible compared to these drives that are only $5 to $50 more. Its not a good long term investment for many reasons.
 
Last edited:
Wow.. how? Mine idles at 40C, loads to almost 60C.... I've never seen these run at 40C... that's idle temps, heatsink or not.
There's two temperature sensors on I assume even the 970 Pro? I know on my 970 EVO the one I selected always has the highest temperature for the second sensor, usually it gets up to 60-65C.
No heatsink
129702
 
Try to get 250 GB of space free and keep it that way if you're concerned about your write speeds. Your 660p's 140 GB SLC cache is probably at 12 GB now with 70%+ of data in there.

Generally, what is really important in day-to-day usage is your read performance, unless you're doing a lot of writes for some reason (e.g. database, video editing cache, etc.).
 
I have over 350gb free space as you can see her but still SEQ is shitty
2019-08-21 12_43_41-Anvil's Storage Utilities 1.1.0 (2014-January-1).png



2019-08-21 12_40_13-AS SSD Benchmark 2.0.6694.23026.png


154gb more free space didn't change anything in AS SSD Benchmark (the last picture)
 
It might be a problem from Windows update attached is a crsytal run of 2 512GB SX8200 Pros in RAID 0. The scores at down by at least 4000 MB/s even now the writes are abysmal for a NVME RAID 0 drive.

-----------------------------------------------------------------------
CrystalDiskMark 6.0.2 x64 (C) 2007-2018 hiyohiyo
Crystal Dew World : https://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 bytes/s [SATA/600 = 600,000,000 bytes/s]
* KB = 1000 bytes, KiB = 1024 bytes

Sequential Read (Q= 32,T= 1) : 1946.712 MB/s
Sequential Write (Q= 32,T= 1) : 1191.809 MB/s
Random Read 4KiB (Q= 8,T= 8) : 374.238 MB/s [ 91366.7 IOPS]
Random Write 4KiB (Q= 8,T= 8) : 1061.326 MB/s [ 259112.8 IOPS]
Random Read 4KiB (Q= 32,T= 1) : 323.136 MB/s [ 78890.6 IOPS]
Random Write 4KiB (Q= 32,T= 1) : 250.047 MB/s [ 61046.6 IOPS]
Random Read 4KiB (Q= 1,T= 1) : 30.350 MB/s [ 7409.7 IOPS]
Random Write 4KiB (Q= 1,T= 1) : 157.868 MB/s [ 38542.0 IOPS]

Test : 100 MiB [S: 7.3% (69.8/953.8 GiB)] (x5) [Interval=5 sec]
Date : 2019/08/21 16:56:09
OS : Windows 10 Professional [10.0 Build 18362] (x64)
 
what a garbage of a drive, never ever buy QLC and especialy those cheap units.
Intel 660p SSDs are actually pretty good drives for the money. So tell us, what was your problem with them.
 
Updated to latest AS SSD Benchmark , what a difference, stille don't explaine the low score i ATTO

2019-08-21 12_40_13-AS SSD Benchmark 2.0.6694.23026.png
2019-08-22 10_45_41-AS SSD Benchmark 2.0.6821.41776.png


There was also a new version of atto but 12mb to 64mb is still low

2019-08-21 12_35_56-Untitled - ATTO Disk Benchmark.png
2019-08-22 11_36_41-Untitled - ATTO Disk Benchmark 4.01.0f1.png
 
Last edited:
That moving point where the performance drops is curious. I wonder what is the actual amount of data ATTO writes to disk. Filling up SLC cache still seems like the most obvious cause here.
Try running ATTO from 8MB to 64MB, will there still be slow results?

I tried running ATTO on my 2TB 660p and while the results were less stable (64KB-2MB ran 1.9GB/s, 2MB-64MB ran 1.5GB/s) there was no real drop like that. Granted, 2TB version has more SLC cache allocated and my drive is currently only about 55% full.

There's two temperature sensors on I assume even the 970 Pro? I know on my 970 EVO the one I selected always has the highest temperature for the second sensor, usually it gets up to 60-65C.
Samsung drives (I have seen this on 950 Pro and 960 Pro and I would assume it is same on all their NVMe drives) have two sensors - one is for drive/Flash and the other one is for the controller. The higher one should be controller.
 
Last edited:
Back
Top