Latest updates from China on prototypes, deliveries and more!


Hi Dudes!

I see that discussion about SSD got hot here :slight_smile:
So lets me clarify some stuff here.

First of all we unfortunately don’t get any special discount from Intel by using Intel SSD. We are not big enough to qualify for special pricing on SSDs.

What we can get is engineering support. It is quite important as you know big part of hardware performance comes from firmware. :slight_smile:

Also I would like to clarify that we were able to switch to 2280 m.2 form factor some time ago. It took us a while but we redesigned the motherboard to fit it in.

A lot of community members before wanted to have PCIe SSD and Intel 600p is a great choice when it comes down to power efficient PCIe SSDs. Big advantage to us would be that Intel could help us tweak the system to get the max performance from their SSDs.

We will measure up power consumption of Intel 600p again on idle. It should be 40 mW which is very similar to Evo.

The main issue with SSDs right now is that there is a big Wafer shortage. Wafer is a raw material used for RAM, SSD and other storage solutions. The reason for it is because smartphones are switchingnto higher RAM capacities as well as storage while more and more PC manufacturers switch to SSD from HDD. So the reality is that getting high end SSDs is very tough. That’s why many pc vendors including Microsoft don’t publicly say what SSD they use and change it time to time based on supply sittuation. Here is the link:

So here is what we will do. We will retest V ourselves and reach out Intel technical support team. Then we will be able to confirm SSD idle power consumption.

Meanwhile we will check availability of 850 Evo. One question to you guys are there any good other PCIe SSDs(850 is not PCIe but we will still check) ?

In general right now we are trying to be as transparent as possible with process. We got a lot to learn for the next project :slight_smile: definitely more prototypes shipped more often to members and making sure we have better time buffers. Now we learned a lot during V creation and it will be one hell lot of a device :slight_smile:


In general. Once Vs are perfected and shipped! Before we will kickoff new project we will. Spend some time together with you community understanding what can be and should be improved in developmebt process. Next time we will have much more structured flow I think :slight_smile:


This might be too late, but could you consider embedding/soldering the SSD chip onto the mainboard?

Samsung released BGA SSD last year, the PM971, followed by Toshiba a few months later with the BG series SSDs. If my approximation is correct, this would buy us around 4 Wh of battery, basically giving us 52 Wh battery or roughly 50 minutes of battery life. If its too late, at least this would be useful for the second-generation V next year :wink:

Here’s the size comparison

One device I know that uses this kind of SSD is the MacBook 12", they were able to basically fit the entire computer in a board as big as a Raspberry Pi, although with no upgradeability whatsoever.

I am right now using the SM951 (similar to 950 Pro) on my desktop PC. While the speed is impressive, I read from the reviews that the power consumption isn’t that great.

One that impressed me is this Sandisk Z400s, where its tested to get better efficiency than the 850 EVO. Apparently the key here is to get a DRAM-less and cache-less SSD. The good news is the pricing is on the lower side, and it seems to be made for OEMs in mind. The bad news is its not PCIe / NVMe, but I stand by the original consensus that NVMe doesn’t provide any significant advantage on this type of devices.

Based on the testing done by TomsHardware, the Z400s is able to get 668 minutes, while the 850 EVO can only manage 621 minutes.

TomsHardware states that they use the Lenovo T440 with 6-cell battery for SATA SSD reviews, which has 47 Wh capacity. (source:,4058.html#p11,

That means, the average power consumption with the Z400s would be:
47 Wh / 668 minutes = 4.22 W
And the average with the 850 EVO:
47 Wh / 621 minutes = 4.54 W

That’s a delta of 0.32 W

If we subtract that to the 4.364W that we got earlier, that would give us:
4.364W - 0.32 W = 4.044 W

That leads to a battery runtime on the Eve V of:
48 Wh / 4.044 W = 11 minutes and 52 minutes.

That is almost an hour difference, which would theoretically increase the claimed battery life from 10-12 hours to 11-13 hours.

tldr: Theoretically, the SanDisk Z400s would let the V get about an hour longer battery life over the 850 EVO, or almost 2 hours over the 600p


Good stuff Patrick.
As for the onboard memory solution we wish we could do it but you gotta be Apple to do it well. This is something we will definitely consider next time.

Thanks for the tip on SSD we will check that one too

@Patrick_Hermawan remeber SSD effeciency will depend a lot on how well it is implemented in the system. So it’s a lot about how well our engineers can optimize how SSD uses power in our device. That’s where Intel support is helpful :slight_smile:

Are there any PCIe SSD s that would be better alternative than Intel 600p while in the same price range?


WOW! If that is the case and there are no negative aspects (e.g. doubling the boot time, significantly slower speeds) I would recommend to strongly look into this. Working together with Intel but not bringing any perks (such as cheaper rates) allows an open observation of the market and thus the 600P doesn’t seem to be the best available option (anymore) - particularly after switching to a more widely used form factor.
(The statement above is made without reviewing the Z400 further myself).


There isn’t many PCIe / NVMe SSD in that price range, but I would recommend taking a look at the Samsung PM961 (similar to 960 Evo). It is about 10% more expensive in the retail market (EU), while providing twice the performance.

Take a look at review here,4737-2.html

While the efficiency isn’t as impressive as Z400s or 850 EVO, it is still slightly better than the 600p, although I didn’t take Intel optimization into account.

EDIT: Other NVMe SSD to consider is ADATA XPG SX8000. Its priced competitively against the 600p in the retail market (US), and they even share the same controller as @iKirin mentioned earlier. It is claimed to use 0.14W slumber and 0.33W typical. Not quite sure what that means, but in term of speed it seems to be faster than the 600p. Data sheet:


It might be possible that if they use an SSD not made by Intel, they can’t use Intel’s optimization anymore… So the battery life would be the same. Just a guess…


Of course this might be possible. We don’t know. So it’s for Eve to judge what benefits they want for the community. But The tests sound like that tomshardware simply put the different SSDs into the same laptop (how else would you achieve comparitive results?) which means that without improvements there is already a longer battery life.
All I was trying to say is that the specs of the 600P that are shared do not seem to be impressive compared to other specs and using a form factor with better availability should allow for open discussion (unless SSDs are already procured, which I don’t see as it was mentioned that after tests of power consumption the decision will be made).


Yes and I’m very appreciative of eve adding this. While its good,a constant reminder (an indicator light) would still be my preference.
It sounds like that’s not possible so week need to wait for v2 got the next round of upgrades. :slight_smile:


Well, it has better battery life without improvements, but if Eve is able to get improvements for the Intel SSD, it might catch up with the power efficiency. Just saying that those benchmarks should be taken with a grain of salt :slight_smile: Of course there is place for an open discussion, I was just dropping my 2 cents into it :slight_smile:


I did quite a bit of research on DRAM-less SSD, and I found out that it is quite harmful for the flash memory itself due to write amplification.


Basically a NAND flash requires a certain minimum of data to be written to each block. As an example, lets say the minimum is 8 KB. If you want to write 1 KB data to the block, and there is an existing 7 KB data on the block, the existing 7 KB data has to be read, added with the 1 KB data that you’re going to write, and then the whole block will be erased and written with the 8KB data. That means, every sub-8KB write wears the SSD out as much as a 8 KB write.

This is where DRAM comes into play. DRAM caches the writes and flush it as one sequential large write. Back to the 8 KB example, we could replace 8 times 1 KB writes (that wears the SSD by 8 KB each) with a single 8 KB write. This reduces the wear by 8x.

Both the Z400s and 600p are DRAM-less, so I suspect they are prone to this issue.

The problem is DRAM cache itself, like the system RAM, uses quite a lot of power, and it can easily tax an hour of battery life on our V. In fact, TomsHardware stated that the additional battery life advantage of Z400s almost single-handedly comes from the lack of DRAM. I dont know why the 600p, while DRAM-less, still consumes a lot of power, but I digress.

Basically here’s the situation:

  • DRAM-less is harmful for SSD’s lifetime (and performance)
  • DRAM uses a lot of power

I got a theoretical solution.

Parts of the system RAM can be allocated as cache for the SSD, essentially acting as the DRAM cache. That means, the 8 GB of system RAM will be partitioned as, say, 7.75 GB for the system, and 256 MB for the SSD. That’s inline with most DRAM-equipped SSDs. (source: Personally, I don’t mind having 7.75 GB RAM instead of 8 GB on my V, especially if it gives me an extra hour of battery life.

Oh, and cache always gives extra performance. Basically we could get away with the lower cost and power consumption of DRAM-less SSD without the disadvantages.


600p is DRAM-less but uses an SLC cache, which theoretically functions like the DRAM on DRAM-SSDs,
without issues with data loss on power-cuts what-so-ever.

The SLC cache is extremely fast especially when use with fast interfaces like PCI-E + NVMe.
But the downside is there more chips to power, more physical delete-write cycles involved.

As what Anandtech has mention, this design is pretty sound on paper, but for some reason the performance does not live up to the specifications. My _un_educated guess is that the 3D TLC they used is slow :confused:

It’s also worth to mention that while SSDs with DRAM is susceptible to data loss upon sudden power cut,
it should have little to do with the V or any other tablets since it virtually never cuts power because of the battery.


Considering of the choice of SSDs to use on the V…

I would consider the 850 EVO as a very fast and power-efficient choice. Anything faster than the 850 EVO is unnecessarily too fast.

I’ve order the 512GB i7 variant of the V, if we’d use the Intel 600p, it’ll have a 17.5GB SLC Cache.
For most of my foreseeable usages, apart from large file copies, the Intel 600p would likely bring me more performance than the 850 EVO because I rarely deal with data set larger than a few GBs in a short period of time. (I copy large files this from time to time for various reasons, mostly with incompressible data, so Sandforce SSDs are not good for me)

While the performance is much appreciated, I would rather have a slightly slower device with better battery life.

TL;DR for the following spoiler: I have a slower SSD that in most cases feels the same as the faster SSD.

[details=Summary]I personally use the venerable Crucial m4 (128GB) on my 9yo laptop. I had been using this SSD for the past 5 or 6 years, initially for my desktop until I replaced it a 256GB Plextor M6S, and not long year, a 1TB Samsung 850 EVO.

While the M4 was a reasonably fast SSD in 2012 standards, it’s more like the budget SSDs today and slow compared to SSDs like 850 EVO and M6S (which is also old now, but at least uses fast MLC). However, it’s rarely the difference of the speeds these SSDs that affect the overall experience.[/details]

Which is why I think extra battery life is more important than extra performance.


Hi @Konstantinos and @Mike,

i’m currently saving money to get my V once you open the store but I wanted to say:

Please avoid Paypal at all costs. What happened to you is sadly very common.
PP It’s simply not right for your volume and it eats into your margin.
Go with Stripe.
Paypal is an arbitrary company, and they operate in the shadows for their own benefit.
They won’t change.
Under the pretense of security they are using you as a bank illegally.
So sue them and get compensation if you can.

Let’s wait until we get the tests results and reviews.
But I’d say for the next revision we should go for battery life.
We already have sacrificed CPU for battery, and 1 extra hour is very significant.
More so now that we still don’t have final battery life results.
SSD performance matters at boot and on the TB scale operations.

3-The best thing so far for me as a future customer is to see you work: your commitment, your transparency, the relentless pursuit of quality (literally at all costs).

The V is one of your firsts products, but for me your values are the hallmark of a great company in the long term.

I’s really great to see you work.

Thank you :slight_smile:


You guys are doing great. I realize there have been delays but giant companies have had delays like that and you guys are managing to keep up with the likes of them! So kudos to you guys for all your hard work, it feels so great to be a part of this community! Look forward to participating in the next project as well


Thanks! So happy to be part of the team :blush:


Correct me if im wrong, but the SLC, being a NAND flash, still suffers from write amplification.

I think if we want to focus on battery life, we need to eliminate as much caches as we can, DRAM or SLC, as long as it doesnt impact the performance by a significant margin.


Your points are well based on the technical knowledge you have :slight_smile: As such I’m happy to ask you: What would be your recommendation? :slight_smile:


I agree that it most probably does suffers from write amplification, since that’s how NAND flash work.

The following is based off the limited knowledge of SSD and NAND flash I have, please also correct me if I’m wrong.

The SLC is a cache, or more accurately in the Intel 600p’s case, a write buffer.
(It isn’t a tiered storage cache like the ones you find in hybrid drives)

Depending on how the SSD controller controls the cache, it theoretically should experience less WA than when used as a permanent storage instead. For example, the write buffer could be immediately emptied once all data has been written onto the TLCs.
(Also recall that for NAND flash, erasing is the slowest operation; and to overwrite, it must first erase.)

WA does not happen when there’s nothing in the pages the controller wants to write data in.
So at least for lighter workloads, WA should be low for the SLC cache.
(E.g. the controller can choose empty pages to write data in)

We should also consider that although the 600p is power hungry when compared to SATA SSDs,
when we compare it to NVMe PCI-E SSDs, it’s still one of the most energy efficient.
This hints that SLC caches probably isn’t the root cause of being power hungry, but with the NVMe PCI-E stuff.


What are the limitations besides the physical size?

For the community to suggest viable alternatives, we need to have a clear understanding of requirements.

For instance, do we still have the option of SATA vs PCIe? If PCIe us it a requirement to have NVMe?

Our putting it another way, are the Samsung 850 (and other similar ssd) option?

Final question, what is the anticipated implementation time frame? Are we looking at first batch, afford batch or Q3?

Thanks fit giving us the opportunity to comment and provide input