• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Skylake Clockspeeds and benchmarks!

Expected Sep-Oct 2015.

"Later this year Intel plans to introduce the true successors of the “Devil’s Canyon” central processing units. The Core i7-6700K and the Core i5-6600K based on the “Skylake” micro-architecture are expected to hit the market in late September or October."
 
September/October is a bit later than I hoped it would be (will still be waiting though), but it makes sense that they wouldn't want to release skylake at the same time as broadwell.
 
How long after this release will we need to wait for their 'tock'? I'm saving a big upgrade for the 16/14nm GPUs in 2016, I'm hoping it will coincide.
 
How long after this release will we need to wait for their 'tock'? I'm saving a big upgrade for the 16/14nm GPUs in 2016, I'm hoping it will coincide.

Skylake is a tock, so I'd imagine it would be at least 2 years until the next 'tock'.

Haswell for desktop was released June 2013, Skylake should be August 2015, so 2 years and 2 months between the most recent tocks (assuming Skylake does release in August).

If by some miraculous event Zen manages to release on time in 2016 and is superior to Intel's current CPU's, we may see a faster tock, though I highly doubt Zen will prove to be the product we all hope for.
 
I think the next tock will be in 2018, as Cannonlake was pushed to 2017.

I wouldn't be surprised if the whole industry stalls and a silicon replacement has to be brought in early and the next shrink ends up being 2020.
 
why no pcie 4 ? pcie 4 with skylake refresh or something ?

Do people not get that no one will implement PCIe 4 until, you know, PCIe 4 is actually finalised, which isn't expected till the end of 2016.

Standards take time and releasing a motherboard with a potential set of final specs on the standard then finding the specs change and your motherboard/chipset is the only one that doesn't actually support it..... well, you're screwed.

There is a very good reason that AMD/Intel/no one else will implement PCIe 4 till post finalised specification. Expect PCIe 4 to appear on new parts from 2017 onwards.
 
Do people not get that no one will implement PCIe 4 until, you know, PCIe 4 is actually finalised, which isn't expected till the end of 2016.

That and we haven't even got any cards yet that can saturate PCI-E 2.0, so it's a bit pointless right now (apart from marketing and separating fools from their money ofc).
 
That and we haven't even got any cards yet that can saturate PCI-E 2.0, so it's a bit pointless right now (apart from marketing and separating fools from their money ofc).

PCI-E SSD's, such as the M.2 Samsung SM951 and the Intel 750 series take full advantage of PCI-E v3. They get absolutely cripped if you limit them to PCI-E v2.

GPU wise, there is a difference between PCI-E revisions in some games, even if using a single powerful GPU. 2way SLI and above obviously show a larger difference between PCI-E speeds.

R0HPIL7.gif.png

This gap will only increase in time, as newer GPU's get even faster etc.
 
It's not true that bandwidth won't increase performance, it's not about saturating the bus. 1GB of data will transfer in half the time on a bus twice as fast. 5ns here and 5ns there add up to delays, shaders sitting doing nothing. It's not unimportant but those results are also very very exaggerated.

http://www.techpowerup.com/reviews/NVIDIA/GTX_980_PCI-Express_Scaling/16.html

Shadow of Mordor has a less than 1% difference going from max speed tp PCIe 3 8x or PCIE 2 16x. The performance summary of for a change for Techpowerup, a fairly new set of games, suggests Ryse is a massive outlier. Though one difference between most of those games and Ryse is it's one of if not the only game not also on last generation consoles. Mordor is on the 360, Watchdogs, a lot of the other games.

Is Ryse a general move towards better utilisation based on not being designed in any way shape or form to run on the last gen consoles.... possibly.

DX12 could also be a push forwards, more compute that requires more data to be sent back to the CPU then more communication and less latency helps. DX12 and more flexibility for the devs and potentially more things done on compute could push games that way.

In terms of SSD performance over PCIe, meh, potential latency reductions going through NVMe instead of ACPI is the biggest improvement for most people. Huge huge sequential performance is of limited use to most people, guys running stupid high resolution raw video footage and working on it maybe. Gaming, meh. We're still heavily reliant on latency and 4k random read performance. The NVMe interface improves latency noticeably over ACPI, but the bandwidth doesn't realistically help all that much. 500MB/s, 1.5GB/s or 5GB/s sequential speed doesn't make that much difference when you have 25-30MB/s random 4k reads for all three.
 
500MB/s, 1.5GB/s or 5GB/s sequential speed doesn't make that much difference when you have 25-30MB/s random 4k reads for all three.

25-30MB for random 4K reads? Nope:

IIV63VZ.png

For typical client workloads, the vast majority of the data use is sequential reads, where these PCI-E V3 SSD's truly are x10 faster than any SATA3 drive.

We should make another thread to continue this, or we'll get way off topic :D
 
Home usage/gaming usage, QD1 is typical and they are pretty much around the 30MB/s mark. Not surprising either, ultimately a single queue(or user) asking for data is requesting a 4kb read then getting it then requesting another one. So the latency is the biggest hold up in the chain. read-request-read-request. NVMe helps reduce each request which increases the number of read/requests.

With higher queue depth you have the latency of requests overlapping and get lots of reads going on. The higher the queue depth the more likely you can get constant reads happening.

It's a shame really, almost every review site has moved over to testing mainly server loads on any storage because it shows up performance differences. Almost no one does a benchmark of how fast a level of a current game loads, something they all used to do. Majority of say Anandtech or most reviews site viewers are home users. It's all how will this ssd effect your home usage, let us show you by informing you how much faster this ssd is over your current one in a server situation in which thousands of requests come in to load a 5MB webpage every few seconds. It's 30% faster.... thus your game will clearly also load 30% faster.

It's not surprising, if Anandtech showed Crysis 3 loading within 0.2 seconds on every SSD available in the past 3 years... they'd stop getting sent free stuff to test.


http://www.anandtech.com/show/8979/samsung-sm951-512-gb-review/7

look at the graph below the one you posted on that page. QD1 is improved on the 951... but not much, it's firmly below the 50MB/s and almost every drive ballpark the same.
 
Home usage/gaming usage, QD1 is typical and they are pretty much around the 30MB/s mark. Not surprising either, ultimately a single queue(or user) asking for data is requesting a 4kb read then getting it then requesting another one. So the latency is the biggest hold up in the chain. read-request-read-request. NVMe helps reduce each request which increases the number of read/requests.

With higher queue depth you have the latency of requests overlapping and get lots of reads going on. The higher the queue depth the more likely you can get constant reads happening.

It's a shame really, almost every review site has moved over to testing mainly server loads on any storage because it shows up performance differences. Almost no one does a benchmark of how fast a level of a current game loads, something they all used to do. Majority of say Anandtech or most reviews site viewers are home users. It's all how will this ssd effect your home usage, let us show you by informing you how much faster this ssd is over your current one in a server situation in which thousands of requests come in to load a 5MB webpage every few seconds. It's 30% faster.... thus your game will clearly also load 30% faster.

It's not surprising, if Anandtech showed Crysis 3 loading within 0.2 seconds on every SSD available in the past 3 years... they'd stop getting sent free stuff to test.


http://www.anandtech.com/show/8979/samsung-sm951-512-gb-review/7

look at the graph below the one you posted on that page. QD1 is improved on the 951... but not much, it's firmly below the 50MB/s and almost every drive ballpark the same.

Yeh I'm not advocating a PCI-E SSD solely to speed up game loading, but for an OS drive, where it would make a massive difference, especially over someone using SATA2 at the moment.
 
Back
Top Bottom