What determines NT8 Playback - Market Replay MAX setting?

[sledge]

I have 2 tests of the exact same thing in NT8, and my powerful desktop is getting crushed. I even set NT8 app processor priority to Above Normal.


I tried IMDISK and moved the Market Replay to memory... it hasn't helped.

So NT8 is assuming something about my system and limiting me. It's like it is setting max maybe to 1500x on the desktop when the laptop is getting 3000x.

The only thing different with the laptop that's special is Raid 0 on the 2 ssd's... my desktop has more cores (6 vs 4) /faster processor speed (3.3ghz vs 2.6ghz)/faster memory (2400mhz 48gb ddr4 vs 1666mhz 16gb ddr3l)

I'd really like to get my desktop as fast, if not faster.

Are points given for M.2 raid 0 over samsung 850 pro ssd SATA 6gbs?
Is Quadro 2100m considered for more replay speed over GeForce 780 ti gtx?

What am I missing?

[MarketAlly]

I have found the SSDs to make considerable difference above SATA. As for the processors, verify the performance between them on https://www.cpubenchmark.net/ to confirm. Last resort, use passmark to see where the bottlenecks are.

[sledge]

I have found the SSDs to make considerable difference above SATA. As for the processors, verify the performance between them on https://www.cpubenchmark.net/ to confirm. Last resort, use passmark to see where the bottlenecks are.

I7 5820k x99 (3.3-3.6ghz) vs i7 4720hq hm87 (2.6-3.4 ghz)

http://us.msi.com/product/workstatio...ion&sku_no=371 (4k version,shipped with upgraded processor).

Your right, should run some benchmarks now, but there is definitely a formula behind it limiting.

[sledge]

So no comment from NT? Nice.

Will that m.2 x4 @ 32gb/s be a worthwhile upgrade?

[NinjaTrader_Brett]

Sorry sledge working through higher then normal regular support volume and apologize for the delay on the beta forum.

For Playback MAX setting it literally "goes" as fast as it can.

What does this mean exactly:

* Reading data from the HDD (I/O Limited)
* Throwing the data to all the NT8 processing threads
* Waiting for the processing threads to finish processing the data
* Reading the next chunk of data and do the whole process again


Meaning there can be a bottleneck at any point in the chain (I/O, CPU, and lastly GPU). As to a GPU bottleneck I highly suspect that and haven't seen it in our internal tests. The primary bottleneck was almost always I/O related and CPU related.

How many seconds can you process per realtime second with Max playback currently?

[sledge]

How many seconds can you process per realtime second with Max playback currently?

Hard to quantify exactly, it's mostly time of day related and season related. But it looks like on the desktop Between 30 and 45 minutes depending on each jump.

Well this at least clears up my concern there was some sorry of formula that limited things.

You've given me some more ideas to test.

I'll move replay data to the 1tb harddisk on laptop and see how much of a slow down there is.. also try imdisk there.

I'll kill all the charts on the desktop to eliminate that processing and see if there is a speed up difference there.


The 780ti isn't reporting much gpu usage. Nothing out of the ordinary like running COD.

[NinjaTrader_Brett]

~30 to 45 minutes per about real world second? Thats pretty good I gotta say.