[nextpage title=”Introduction”]
ECS calls “SDGE” (Scalable Dual Graphic Engines) its motherboards that have two PCI Express x16 slots. These motherboards, however, can have three different architectures to host these two PCI Express x16 slots:
- Using SLI or Crossfire technology. In this case, each slot works at 8x when running under SLI or Crossfire modes.
- Using spare PCI Express x1 lanes to connect the second PCI Express x16 slto. While the first PCI Express x16 slot really runs at 16x speed, the second slot doesn’t run at 16x. Its maximum speed will depend on how many lanes were used to connect this slot to the chipset. Besides having two x16 slots, these motherboards cannot run under SLI nor Crossfire modes.
- A new prototype presented by ECS where the two slots really run at 16x speed and where it is possible to have two SLI or two Crossfire video cards together in a system without using a chipset from NVIDIA or ATI. For example, a motherboard with SDGE technology based on an Intel chipset will allow you to run two SLI VGAs in parallel. In this article we will be discussing this third option, which is still a prototype. So all our references to “SDGE” within this article refer to this prototype and not to the other two architectures.
Basically SDGE motherboard will have two 16x PCI Express slots. These two slots are really x16 and not 8x as it happens on SLI motherboards when you enable SLI mode. To make these slots to work as SLI or Crossfire, however, ECS will need both NVIDIA and ATI to unlock their drivers to enable SDGE. This will be a very though job, but ECS said “we are working on that”. Let’s wait and see.
Figure 1: SDGE technology overview.
In our opinion, this concept is very interesting if the user didn’t decide which way to go yet, ATI or NVIDIA. You can use two SLI cards from NVIDIA today and, in the future, change to two Crossfire cards from ATI without needing to replace the motherboard. Really interesting – but, as mentioned, unless ECS gets the unlocked drivers from both ATI and NVIDIA, this concept is useless.
Also, since the second x16 PCI Express slot is only activated through an add-on card, as we will explain in the next page, the motherboard can be cheaper, which can be a very good solution for users on budget: you can buy a “normal” motherboard now and “upgrade” it to have SLI or Crossfire function later by installing an add-on card.
[nextpage title=”How It Works”]
SDGE motherboards have two PCI Express x16 slots, but one of them is not operational. To enable the second slot, you need to install an add-on card (SDGE card). This add-on card has a chip (HyperTransport tunnel chip) that enables it.
The idea behind this strange concept is to provide a cheap motherboard that you can transform it, in the future, into a SLI or Crossfire motherboard.
Basically, SDGE connects the extra 16x PCI Express slot on the HyperTransport bus. Since the location of this bus on the system depends on the platform, the architecture of SDGE varies depending on the platform. Let’s take a look at how SDGE is built on both AMD and Intel platforms.
ECS gave us a Flash animation showing how this technology works. Click here to watch this animation.
AMD Platform
Since socket 754 and socket 939 processors use the HyperTransport bus to communicate with the north bridge, the SDGE card is installed between the CPU and the north bridge chip, as you can see in Figure 2.
Figure 2: How SDGE works on AMD platform.
The HyperTransport Tunnel chip located on the SDGE card connects the extra 16x PCI Express slot on the HyperTransport bus. The original 16x PCI Express slot continues to be connected to the north bridge chip. As we mentioned both slots work at 16x rate.
Intel Platform
Intel CPUs don’t use the HyperTransport bus. However, the link between the north bridge (or MCH, Memory Controller Hub, as Intel calls their north bridge chips) and the south bridge (or ICH, I/O Controller Hub, as Intel calls their south bridge chips) use the HyperTransport bus. So on Intel platform SDGE card is connected between the north bridge and the south bridge link, as you can see in Figure 3.
Figure 3: How SDGE works on Intel platform.
The funny thing is that, as far as we know, only NVIDIA and ULi chipsets use the HyperTransport bus to connect the north bridge to the south bridge. So, for Intel platform, this technology would be available only on motherboards using chipsets from NVIDIA or ULi – and not for all chipsets like ECS is claiming. Since this is only a concept, we still need to wait for the final product in order to see what they’ve done. Click here to learn more about this subject.
The HyperTransport Tunnel chip located on the SDGE card connects the extra 16x PCI Express slot on the HyperTransport bus present between the north bridge and the south bridge. The original 16x PCI Express slot continues to be connected to the north bridge chip. As we mentioned both slots work at 16x rate.
[nextpage title=”Motherboards Using SDGE Technology”]
We must be very careful since ECS is calling all their motherboard with two PCI Express x16 as “SDGE” and this is causing a lot of confusion, especially because they didn’t release any motherboard with the add-on module we described yet.
ECS is calling their SLI and Crossfire motherboards as “SDGE”, even when they don’t have the describe module, only because they have two PCI Express x16 slots – actually these slots run at 8x speed when SLI or Crossfire mode is enabled.
Worst is the case of the motherboards that don’t have these technologies and are also being called “SDGE”, like PF5 Extreme, which is based on Intel 945P chipset. Since this motherboard doesn’t use the described architecture, the second PCI Express x16 slot is made by using spare PCI Express x1 lanes, hence the second x16 PCI Express slot can only run at best at 6x, since Intel 945P chipset has only 6 PCI Express x1 lanes. This is so true that when two video cards are used on this motherboard its PCI Express x1 slot cannot be used.
So, for ECS “SDGE” means “motherboard with capacity of having two PCI Express video cards at the same time”.
We leave here a construtive critic to ECS: Why not using different names? Using the same name to define three distinct architecture causes nothing but confusion.
Leave a Reply