• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Razer Core Ship Dates

So this is just an external box to connect a desktop gpu to a laptop? Seems pretty overpriced at $500 if so.
 
Samson4EiT said:
So this is just an external box to connect a desktop gpu to a laptop? Seems pretty overpriced at $500 if so.
Click to expand...

Yeah, I must be missing something here. You can build this yourself for $50. Surely Razer is offering something more than a metal chassis and a built-in USB 3.0 hub for your extra $450 right?
 
CEpeep said:
Yeah, I must be missing something here. You can build this yourself for $50. Surely Razer is offering something more than a metal chassis and a built-in USB 3.0 hub for your extra $450 right?
Click to expand...

I'd love to see your parts list to build this for $50. You don't get to skimp out, I want to see full TB3 implementation with a USB-C connector. Extra USB 3.1 ports, and an Ethernet jack.

I'll agree that Razer's box is a little pricey, but it's a shit ton nicer than that Devil Box, which ironically is ugly as sin.
 
And they are OOS again. Damn, they must not have much stock!
 
Available for preorder again, shipping date is 12th August 2016
 
Proof once again that external thunderbolt docks are absolutely pointless. You get something resembling a small desktop anyway. And the small market means the cost will be high.

Plus, you get the privilege of castrating your high-end GPU on 1/3 the bus speed of a normal PCIe x16 slot!

By the time you drop that kind of cash, you could have just built the rest of a Core i5 MiniITX gaming system. And then you get your choice in case design!
 
I actually love the idea of these products, just me though apparently. I would love to daily a razer ultra book, then come home after work if I want to game and plug this bad boy in to a 34" monitor. And get reasonable fps. I just wish it didn't have a "u" cpu
 
defaultluser said:
Proof once again that external thunderbolt docks are absolutely pointless. You get something resembling a small desktop anyway. And the small market means the cost will be high.

Plus, you get the privilege of castrating your high-end GPU on 1/3 the bus speed of a normal PCIe x16 slot!

By the time you drop that kind of cash, you could have just built the rest of a Core i5 MiniITX gaming system. And then you get your choice in case design!
Click to expand...

Proof once again that some people are completely oblivious to the reason why others might be interested in such a product. I'm sorry if you're not interested in it, but take the negativity elsewhere.
 
defaultluser said:
Proof once again that external thunderbolt docks are absolutely pointless. You get something resembling a small desktop anyway. And the small market means the cost will be high.

Plus, you get the privilege of castrating your high-end GPU on 1/3 the bus speed of a normal PCIe x16 slot!

By the time you drop that kind of cash, you could have just built the rest of a Core i5 MiniITX gaming system. And then you get your choice in case design!
Click to expand...

So PCIe x16 2.0 (the only one being used) is 8GB/s while Thunderbolt 3.0 is 40Gbit/s which is 5GB/s. I don't know about you, but in my books that's over half.
 
-Strelok- said:
So PCIe x16 2.0 (the only one being used) is 8GB/s while Thunderbolt 3.0 is 40Gbit/s which is 5GB/s. I don't know about you, but in my books that's over half.
Click to expand...

Thunderbolt 3.0 on active cable = 10 Gbps per-lane, actual, with four lanes. That's 40Gbps on a single cable.

PCIe 3.0 = 8 Gbps per-lane. That's real bandwidth, no overhead like PCIe 2.0.

8 Gbps * 16 lanes = 128 Gbps per-slot.

40 Gbps / 128 Gbps = 0.3125 = approximately 1/3 the bandwidth.

I don't know why you insist on comparing PCIe 2.0 speeds against this thing. We've had 3.0-capable cards since GCN and Kepler were released.
 
Last I checked, you only lost ~5fps by cutting the pci-e bandwidth in half on a GPU... Not that big a hit...
 
acquacow said:
Last I checked, you only lost ~5fps by cutting the pci-e bandwidth in half on a GPU... Not that big a hit...
Click to expand...

You're cutting it by a factor of 3, not 2.

There's also the small but notable overhead of transferring the frames back to the monitor over that Thunderbolt link, which can vary in overhead based on your resolution:

1080p, 60Hz, 24-bit framebuffer = 373MB/s = ~3Gbps
4k, 60Hz, 24-bit framebuffer = 1,492 MB/s = ~12Gbps

That's sucking down 30% of your link bandwidth at 4k! And resolutions in that ballpark are quite common on these high-end ultrabooks.

So now we're talking 128Gbps PCIe bandwidth to the card, and Thunderbolt 3 has a real transfer to card of between 28 to 37Gbps! That's now almost down to 1/4 the bandwidth of a dedicated 3.0 x16 slot, worst-case!

All those pretty ultra-high-DPI pixels are your worst nightmare when you're trying to play a game on Thunderbolt. YOU NEED A SUPER-POWERFUL CARD like a GTX 1080, but you don't have the bandwidth to feed it!

I'm just warning people how much of a waste of money this is before they get waist deep into this bullshit. Because at that point, it's your own fault :D

I'm just sick of people buying into these convertibles, and then complaining that they're slower than a desktop Core i5, and the integrated graphics can barely play DOTA. NO SHIT SHERLOCK. For every person like Ruoh who SORTA know what they're doing, there's five who have no idea what they're getting into, and only follow the buzzwords of "Surface is our Savior" and "Praise Be to our all-powerful Thunderbolt."
 
Last edited:
defaultluser said:
Thunderbolt 3.0 on active cable = 10 Gbps per-lane, actual, with four lanes. That's 40Gbps on a single cable.

PCIe 3.0 = 8 Gbps per-lane. That's real bandwidth, no overhead like PCIe 2.0.

8 Gbps * 16 lanes = 128 Gbps per-slot.

40 Gbps / 128 Gbps = 0.3125 = approximately 1/3 the bandwidth.

I don't know why you insist on comparing PCIe 2.0 speeds against this thing. We've had 3.0-capable cards since GCN and Kepler were released.
Click to expand...

Well that was a complete derp on my part. Still, it'll be better than any GPU you can put in a reasonable laptop. I'm not sold on the $500 part though. I'll be waiting a few more years, but I like the idea since GPU's are the most common part to change.
 
-Strelok- said:
Well that was a complete derp on my part. Still, it'll be better than any GPU you can put in a reasonable laptop. I'm not sold on the $500 part though. I'll be waiting a few more years, but I like the idea since GPU's are the most common part to change.
Click to expand...

This is true, but I keep highlighting the fact that this is most of the way toward a Core i5 MiniITX gaming system. Or should be EXACTLY the same price if you switch the CPU with a Core i3:

PCPartPicker part list / Price breakdown by merchant

CPU: Intel Core i3-6100 3.7GHz Dual-Core Processor ($110.99 @ SuperBiiz)
Motherboard: ASRock B150M-ITX Mini ITX LGA1151 Motherboard ($68.99 @ Amazon)
Memory: G.Skill NT Series 16GB (2 x 8GB) DDR4-2133 Memory ($52.99 @ Newegg)
Storage: Sandisk SSD PLUS 240GB 2.5" Solid State Drive ($62.99 @ Amazon)
Case: Fractal Design Node 202 HTPC Case w/450W Power Supply ($129.99 @ SuperBiiz)
Operating System: Microsoft Windows 10 Home OEM 64-bit ($85.86 @ Amazon)
Total: $511.81
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2016-06-27 15:45 EDT-0400

And you don't get any of the insane upcharges you usually get to max-out memory and storage size. You can spend that cash on a more powerful GPU!
 
defaultluser said:
And you don't get any of the insane upcharges you usually get to max-out memory and storage size. You can spend that cash on a more powerful GPU!
Click to expand...

Beating a dead horse. It's not relevant to the thread topic of shipping availability of the product.
 
defaultluser said:
You're cutting it by a factor of 3, not 2.

There's also the small but notable overhead of transferring the frames back to the monitor over that Thunderbolt link, which can vary in overhead based on your resolution:

1080p, 60Hz, 24-bit framebuffer = 373MB/s = ~3Gbps
4k, 60Hz, 24-bit framebuffer = 1,492 MB/s = ~12Gbps

That's sucking down 30% of your link bandwidth at 4k! And resolutions in that ballpark are quite common on these high-end ultrabooks.

So now we're talking 128Gbps PCIe bandwidth to the card, and Thunderbolt 3 has a real transfer to card of between 28 to 37Gbps! That's now almost down to 1/4 the bandwidth of a dedicated 3.0 x16 slot, worst-case!

All those pretty ultra-high-DPI pixels are your worst nightmare when you're trying to play a game on Thunderbolt. YOU NEED A SUPER-POWERFUL CARD like a GTX 1080, but you don't have the bandwidth to feed it!

I'm just warning people how much of a waste of money this is before they get waist deep into this bullshit. Because at that point, it's your own fault :D

I'm just sick of people buying into these convertibles, and then complaining that they're slower than a desktop Core i5, and the integrated graphics can barely play DOTA. NO SHIT SHERLOCK. For every person like Ruoh who SORTA know what they're doing, there's five who have no idea what they're getting into, and only follow the buzzwords of "Surface is our Savior" and "Praise Be to our all-powerful Thunderbolt."
Click to expand...

Read this: PCIe 3.0 x8 vs. x16: Does It Impact GPU Performance?
And this: Impact of PCI-E Speed on Gaming Performance

I was wrong in my original ~5fps estimation, it's not even really 1fps...

I think you are over-estimating how much data the GPU actually pulls from the link... the only thing feeding it really are possibly ssds for texture loading... most all of the processing is internal and doesn't care about link speed.

Video data isn't going over that link, just some textures, map files that get loaded, and maybe some physX stuff...

-- Dave
 
Last edited:
I was originally interested in this, but I gave up waiting and built a PC. Glad I did, because the price is terrible.
 
acquacow said:
Read this: PCIe 3.0 x8 vs. x16: Does It Impact GPU Performance?
And this: Impact of PCI-E Speed on Gaming Performance

I was wrong in my original ~5fps estimation, it's not even really 1fps...

I think you are over-estimating how much data the GPU actually pulls from the link... the only thing feeding it really are possibly ssds for texture loading... most all of the processing is internal and doesn't care about link speed.

Video data isn't going over that link, just some textures, map files that get loaded, and maybe some physX stuff...

-- Dave
Click to expand...

I can't read the Gamers article here, so i'll just comment on Puget.

You're referencing an article from 2013, and still think it's relevant? Also, this article doesn't highlight minimum framerate at all, so it's WORTHLESS from a REAL-WORLD performance impact (i.e. taking frame rate consistency into account).

The GTX 1080 is about 2.25 times the speed of the Titan, and the complexity of compute that has to be sent to the card has grown enormously in those three years. To pretend that this does not require a faster link is just being ignorant of the march of technology.

And once again, after taking into account the link overhead for sending the rendered frame buffer back to the display, you're CUTTING THAT 40GBps BANDWIDTH DOWN to 28Gbps, LESS THAN PCIe 3.0 x4. You're dreaming if you think that won't impact things.
 
Last edited:
defaultluser said:
You're referencing an article from 2013, and still think it's relevant? Also, this article doesn't highlight minimum framerate at all, so it's WORTHLESS from a REAL-WORLD performance impact (i.e. taking frame rate consistency into account).

The GTX 1080 is about 2.25 times the speed of the Titan, and the complexity of compute has grown enormously in those three years. To pretend that this does not require a faster link is just being ignorant of the march of technology.
Click to expand...
Um, yes, it's still relevant... also the first article does talk about minimum framerate...

I've been doing PCI-e storage for 6 years, building the fastest storage systems on the planet... I know how to utilize pci-e bandwidth and I can tell you that GPUs (at least for gaming) don't use much at all.

You've got a 5GB/sec pipe... where do you think you are going to pull 5GB/sec from? Certainly not any hard drives or SSDs in your system. Maybe if you had two pci-e SSDs to feed it with data, but that's certainly over-kill for any games out there (games designed to be loaded off of a hard drive). Your game certainly doesn't require 5GB/sec worth of compute from your CPU...

For reference, the fastest pci-e storage system I built (back in 2010) does 500GB/sec and 50M IOPS/sec...and it's still running today.

-- Dave
 
Okay, now that I can read the article, Im take it back. 30 Gbps is not going to limit things much.

Begs the question though: if we have all this unused bandwidth just sitting here, why did Nvidia upgrade the SLI bridge instead of doing PCIe SLI?
 
defaultluser said:
Okay, now that I can read the article, Im take it back. 30 Gbps is not going to limit things much.

Begs the question though: if we have all this unused bandwidth just sitting here, why did Nvidia upgrade the SLI bridge instead of doing PCIe SLI?
Click to expand...
If you do it through pci-e, you have to involve the CPU and take a chunk of main memory. That's likely slower than them just developing/maintaining their own protocols/methods...
 
Actually the way AMD sold it, they added a dedicated card to card DMA engine, which minimizes CPU impact.

Go read one of the many XDMA articles written about Hawaii.
 
I'm quite read up on RDMA, but that still requires some host memory... definitely better from a CPU-usage perspective.

We worked directly with NVIDIA to get that setup with our ioDrives so that we could attach a few TB of fast pci-e flash directly to the GPUs. The national labs were very interested in it for some of their compute projects.

-- Dave
 
I've decided to just go straight m.2 -> pcie for the time being.
 
defaultluser said:
Actually the way AMD sold it, they added a dedicated card to card DMA engine, which minimizes CPU impact.

Go read one of the many XDMA articles written about Hawaii.
Click to expand...

Having fine control over data transmission and taking PCIe speed out of the question is likely one of the reasons. I remember reading somewhere that it actually still does transfer some data through the bus, but just the lion's share using the bridges. Making it so that the end-user's PCIe generation won't matter is also a consideration.
 
You must log in or register to reply here.
Back
Top