We recently had a chance to sit down with Tom Petersen, the technical marketing manager of the NVIDIA MCP group and find out what's new and have him answer some of the questions that our readers have been asking.
LR: So, what's new?
NV: Wow! That's a loaded question! Seriously though, things have been crazy here! We're working on so many things, there's barely time to breath!
LR: What's so crazy?
NV: It's an interesting time of year. We've got a couple of major trade shows coming up in March-Cebit and GDC-and right after that we'vegot E3 and Computex. All of these shows are a great place to showcase new technologies and products, and trust me, we've got a lot of things in our product pipeline!
LR: What new products?
NV: Well, you know we can never talk about unannounced products. But for those of you lucky (or not lucky) enough to be attending Cebit in early March, you'll see some brand new products on display at the NVIDIA booth, and at many of our partner booths as well.
LR: Can't you give us a hint?
NV: Nope, but given that I work with the NVIDIA MCP business group, I'm sure you can theorize the types of products we will be showing.
LR: New core-logic solutions?
NV: (laughs) Well, that would certainly be an educated guess!
LR: Talking about core-logic, NVIDIA has been on a tear lately. What's the secret of your success?
NV: Well, we are an extremely focused company. When we originally launched the NVIDIA nForce product line about 5 years ago, we were concentrating on the enthusiast market. Obviously, our NVIDIA nForce4 SLI X16 and NVIDIA nForce4 SLI MCPs are the hottest products available today for both AMD and Intel platforms, and we have more than 50% share in this market segment alone.
LR: Why are you only focused on enthusiast platforms?
NV: We're not. In 2005, we launched our latest solution for UMA platforms-the NVIDIA 430/410 family, which paired our award-winning core-logic with our award-winning GeForce GPUs. We launched these products in October and it quickly became one of the fastest product ramps in terms of revenue in the Company's history. We also have a full
family of scalable MCPs for the professional markets-NVIDIA nForce Professional-which are targeted for the AMD Opteron-based workstation and server markets.
LR: How are you doing in those areas?
NV: We are doing great. The adoption rate for NVIDIA nForce Professional MCPs for AMD Opteron-based workstations is about 100%. For AMD Opteron-based servers, a market which is primarily served by AMD chipsets, we currently have about 25% market share. All of the leading workstation and server companies, including HP, IBM, Sun, and FSC, all utilize NVIDIA nForce Professional MCPs. It's a growing business for NVIDIA and you will be hearing a lot more about our growth in these segments throughout 2006.
LR: What about mobile? That seems like a huge market opportunity for NVIDIA.
NV: It's coming. Sooner than you think!
LR: ATI is getting ready to release RD580, and have been critical of your PCIe implementation in your NVIDIA SLI X16 architecture, which splits the dual PCIe x16 lanes between two chips. Care to comment?
NV: Well, we haven't seen RD580, so I couldn't comment on their product and how their products will be perceived in the market. That said, we have noticed that there seems to be a higher level of misinformation about our products floating around recently.
LR: Like what?
NV: Well, for example, there have been reports that the link between our SPP and MCP is limited to an 8x HyperTransport connection. This is not true. The link width is 16x in both directions between the SPP and MCP.
The HyperTransport frequency between both of these links is 1000MHz (5x multiplier).This amounts to a staggering 4GB/sec of bandwidth between the two chips. We have the same link between the SPP and the AMD CPU. Remember, the AMD CPU has a memory controller built-in.
LR: But, doesn't having the PCIe x16 lanes split between two chips impede performance?
NV: Nope. There are no performance issues caused by having the PCI Express X16 slots connected to different chips. NVIDIA SLI graphics performance is predominantly determined by GPU performance and their bandwidth to memory and each other - not latency. Since NVIDIA's X16 architecture provides more than enough bandwidth through its full x16 HT link, there is no appreciable performance difference.
As we said, the NVIDIA two-chip design is comprised of the System Platform Processor (SPP) and the Media and Communications Processor (MCP). The HyperTransport x16 link between them delivers 4GB/sec of bandwidth. This is four times the bandwidth of other chipsets, including those from ATI and Intel, which only use a PCIe x4 or PCIe x2 lane between their two chips. We use HyperTransport. They don't.
LR: Yes, but what about a situation where you would completely utilize all of the bandwidth of the link?
NV: Well, the truth is, you would never ever have a real-world environment where this would happen. Nobody, including Intel, designs chipsets for hypothetical usage models. At no time would you ever encounter a situation where you would have to sustain peak bandwidth for PCIe, SATA, and networking, all at the same time. All chipset designers can do, is to design for real-world usage models. This applies not just to NVIDIA, but to everyone else too, including ATI and Intel.
For example, ATI uses a slow 2x PCIe interconnect between their northbridge and southbridge with their SB400 families, and a 4x PCIe interconnect when they pair their northbridge with the Uli/NV southbridges. In these cases as well, based on a hypothetical usage model, they too would not have sufficient bandwidth to handle all of the southbridge functions (dual GigE, 4 Sata 3GB/sec drives, USB, etc.) The same applies for Intel, which also uses a 4x PCIe interconnect between their two chips.
To address your claim, NVIDIA provides plenty of bandwidth between the SPP and MCP for games. And, outside of games, that link provides plenty of bandwidth for the PCI, SATA, gigabit Ethernet, USB 2.0 and other connectors tied to it. The same cannot be said for the competition.
While hypothetical arguments are fun to engage in, they really have no basis for drawing performance conclusions since they don't relate toreal-world usage models.
LR: Is there an architectural advantage to combining both X16 PCI Express slots on a single chip rather than splitting them between the SPP and MCP as NVIDIA does with the NVIDIA nForce4 SLI X16 architecture?
ANSWER: Not really. The platform is only as good as the design of the overall architecture. HyperTransport is a high-speed, low latency, chip-to-chip interface that is optimized for data throughput. As we mentioned earlier, we provide sufficient bandwidth in the inter-chip HyperTransport link to yield the same effective performance.
One way to think about this intuitively is to understand that the functions accomplished by both architectures are identical. Both single chip and dual-chip designs must queue up PCIe packets at the GPU interfaces and then arbitrate, steer and transfer the packets to the CPU and main memory. It turns out that for an average packet, the transfer latency is dominated by the first 3 functions and actual transfer between the SPP and MCP represents a small fraction of total packet latency.
When you look at the actual architecture of most chipset solutions, the link you really need to look at is the one between the northbridge and CPU. This is where all data eventually has to go anyways, in order to access main memory.
LR: But, we've seen some reviews where you see negative scaling going from NVIDIA nForce4 SLI (dual x8) to the new NVIDIA nForce SLI X16. Why is that?
NV: (laughs) Yes, we have seen those too. Unfortunately, some of those early reviews were conducted with improper BIOS settings. It is imperative that you run with proper BIOS settings to ensure full performance on NVIDIA nForce4 SLI X16 platforms. Running the HyperTransport link at 8x as some have done in comparative reviews unfairly negatively impacts performance. The HyperTransport link width of the NVIDIA nForce4 SLI X16 MCP is x16 in both directions and platforms should be reviewed properly. For those interested, there is measurable scaling between NVIDIA dual X8 and dual X16 platforms with selected applications, most notably when turning on SLI AA modes.
LR: Why does NVIDIA continue to offer SLI technology only on NVIDIA-based platforms?
NV: NVIDIA has an extensive certification program for the entire SLI ecosystem, which now consists of GPUs, motherboards, power supplies, memory, PC cases and full PC systems. This certification program and the
testing of the interoperability of all of these components is very complex and requires many resources. As a result, we have elected to focus our certification program on motherboards featuring NVIDIA nForce MCP technology. Motherboards are available for both Intel and AMD platforms, at a variety of price segments, starting as low as $79 USD.
LR: Why doesn't NVIDIA support HDA?
NV: NVIDIA supports HDA on a variety of nForce products. For the high-end enthusiast market, studies have shown that end-users prefer add-in audio cards rather than on-board audio. HDA is important to the HTPC segments of the market and lower-cost integrated solutions. All of our newest MCPs fully support AC'97 as well as HDA.
LR: Okay, last question. NVIDIA QuadSLI. Isn't this overkill?
NV: We certainly don't think so! The ability to drive ultra high-resolution monitors, crank up the eye candy and still havesilky-smooth frame rates is definitely something we are excited about delivering. It also showcases the ultimate power of the PC and reinforces its position as the ultimate gaming platform.
LR: What will you need to run it?
NV: A motherboard using NVIDIA nForce4 SLI MCP technology of course! That, and the Quad SLI GPUs based on NVIDIA GeForce GPU technology. Really, Quad SLI showcases the beauty of NVIDIA innovation. Since we
have incorporated SLI technology directly into the design of the GPU, only NVIDIA has the ability to offer this level of performance and gaming. We're really excited about it, and look forward to the market reception.
LR: When will we able to get it? And, will it be available for AMD platforms too?
NV: Soon, and you bet!