NVIDIA GeForce GTX 880 “Maxwell” Engineering Sample Analysis – Features 8 GB GDDR5 Memory and 2nd Generation GM2** Core

Hassan Mujtaba
Posted Jul 3, 2014
44Shares
Share Tweet Submit

NVIDIA is quietly working on the GeForce 800 series graphics card which will replace the GeForce 700 series cards and feature the latest Maxwell graphics processing unit. While there have been hints that the GeForce GTX 880 ( the flagship GeForce 800 series card) will launch in Q4 2014, it looks like GamerSky, a Chinese website has leaked the first engineering sample of the GeForce GTX 880 PCB which gives us an early glimpse of what to expect from NVIDIA’s upcoming flagship component.NVIDIA GeForce GTX 880 ES Board_1

NVIDIA GeForce GTX 880 “Maxwell” Engineering Sample Analysis

For starters, the GeForce GTX 880 PCB we are currently looking at is in ES phase and as stated by the source, the GeForce Maxwell core is labeled “1421A1” which is the A1 revision so this is a very early sample, prepared in the month of May (19-25) or the first 21 weeks of 2014. Its possible that we might see variations in the final revision that will use A2 and B1 revisions on the GeForce cards but the most important thing hasn’t been mentioned that whether or not the card features a GM204 core or a GM210 core. But before we go into details regarding what core is featured on the PCB, let’s take a journey in the past and tell you (our readers) why we believe this is a very concrete leak.

If you had been our readers since a long time, you would know that the GeForce GTX 680 PCB which was also leaked almost two years ago had a very similar design scheme. The GeForce GTX 680’s PCB was also in ES phase and featured a green PCB and the same memory configuration as the one on the leaked PCB of the Maxwell card. Compared to the GeForce GTX Titan and 780 Ti PCB, the memory configuration is less dense since the cards make use of a lower memory bus hence the similar arrangement of 8 memory chips on the front and 8 memory chips on the back which amount to a total of 16 memory chips on the entire PCB. The GeForce GTX 680 did feature only 8 memory chips on the front but NVIDIA making a big move on the GeForce GTX 880, equipping it with 8 GB GDDR5 memory was quite unexpected but was rumored earlier. The memory chips in specific are the SKHynix H5GQ4H24MFR 0.5GB which are rated to operate at 7 GHz effective clock speed, similar to many NVIDIA Kepler based cards which would be a nice improvement over the current generation.

Moving onward, we have the GPU core, it is not explicitly stated on the die whether the GeForce Maxwell chip being used is either GM204 or GM210 but the source speculates that it could be the GM210 due to the shear size of the PCB which I don’t believe is a possibility. There are several reasons for this, the GPU size is relatively larger than the GK104/GK114 which were featured on the GeForce GTX 680 and GeForce GTX 770 but smaller than the GeForce Kepler GK110 which is relatively larger. A decent comparison of the GPUs was made by Videocardz which can be seen below:NVIDIA GeForce Maxwell GM204

Now some may believe that the larger die on Maxwell equates to the card using the flagship GM200/GM210 core but that’s not true. As we have already seen, NVIDIA increased the die size of their Maxwell based GM107 which measured 148mm2 compared to its predecessor, the GK107 which measures 118mm2, so more or less this confirms that the flagship Maxwell core will be larger than GK110 in size but more efficient in terms of performance and power. Much of the technical details of the second generation Maxwell core or not known but some hints were given in rumors and previous reports which will be shared at the end of this article. So this just might be the GM204 core with some early phase designs to test stability. The PCB size is something which we have seen this generation with board manufacturers extending the width of the PCBs to add more power components to increase stability, overclocking potential and this is generally what companies do in ES phase.

NVIDIA GeForce GTX 1060 3GB Specs Leaked – Half The Video Memory And More Than A Few Changes Between 6GB Version

The back of the PCB is not much interesting but I happened to find out this relatively surprising marking near on of the screw holds which mentions “256”. Its hard to say what this means but we did hear a 256-bit bus being used on the GeForce GTX 880 and with a similar memory configuration as the GeForce GTX 880 on the front end, this might indeed be a conformation for a 256-bit bus being used on NVIDIA’s latest flagship Maxwell card how-ever, its wise to wait for more official and concrete specs but sometimes, speculation does turn out to be true.

The power area on the PCB is very interesting and once again, we see the GeForce GTX 680 resemblance in the stacked pin connector design. The stacked pin connectors were officially launched with the GeForce GTX 680 but NVIDIA didn’t adopt the design with their other cards, possibly due to the NVTTM being a highly efficient cooling design. These being featured on the PCB might just be an ES design or NVIDIA may actually be bringing this back. The surprising part is that the card makes use of two 6-Pin and a single 8-Pin connector. This does points out to a TDP of 325W which is definitely not something NVIDIA wants to aim for on their next generation GPU design which are focused and marketed for being “highly efficient”. But once again, I believe that this is just a design phase board to test stability and evaluate the card for final qualification testing. The GeForce GTX 880 at most is expected to feature a TDP of around 180-225W which falls in line with their GeForce offerings during the Kepler generation, especially if this is based on GM204.

The main PCB shot is pixelated blurring some very important parts of the PCB from our eyes. There are two massive black blocks at the back of this PCB which have been mentioned to be Denver cores but Denver isn’t as massive as the blocks being shown and they are simply cooling fans used on the PCB to cool the VRM and power components, most of which can be seen in the back shot of the  PCB. The front Display connectors are also blurred but we can see two SLI gold fingers which means that NVIDIA isn’t ditching the good old connector style multi-GPU technology in favor of a AMD XDMA styled design. For cooling, NVIDIA recently patented the so called Turbofan which is a Axial-Radial Hybrid fan design similar to what MSI’s currently offering on their current Mini-ITX cards. So we might see a new cooling design on the card and details on the new axial-radial cooling fan are given below:

“The present invention provides a turbofan and a graphics card with the turbofan. The turbofan comprises: a turbofan assembly which admits air in an axial direction and dispenses air in a radial direction; an inlet fan assembly disposed at an inlet of the turbofan assembly and disposed coaxially with the turbofan assembly: and a driving means for driving the turbofan assembly and the inlet fan assembly to rotate. The turbofan provided by the invention gathers the ambient air to the inlet through the inlet fan assembly disposed at the inlet of the turbofan assembly, so as to change a negative pressure state at the inlet. Consequently, the cooling efficiency of the turbofan is improved effectively and the noise of the turbofan is reduced.” via BitsnChips

Price wise, the GeForce 800 series would stick with the $499 US price for the flagship and $399 US price for the cut-down variant as with the first generation Kepler parts. NVIDIA will not adopt the $649 or $549 US pricing which they went with the GK110 series and performance wise, the card would be better in both performance additionally, the power consumption. You can derive more technical details of the GeForce Maxwell architecture from the GM107 details but it should be noted that the GeForce GTX 880 will possibly make use of the second generation Maxwell core. Specifications and additional details on Maxwell GeForce cards are listed below:

NVIDIA Unveils The GeForce GTX 1060 6 GB For $249 US - Features GP106 Core With 1280 Cores, 1.7 GHz Core and 8 GHz Memory Clock, 19th July Launch

NVIDIA Maxwell GM107 Architecture

The NVIDIA Maxwell GM107 architecture is most clearly built from scratch yet looks like an hybrid of both Kepler and Fermi. The SMM or Streaming Multiprocessor of Maxwell will replace the SMX of Kepler and each of the smm are assembled into four blocks yet are defined as part of a single SMM which means the core architecture has got quite dense with Maxwell. Each of these blocks hold 32 CUDA cores so a single SMM with four of these operation units results in 128 CUDA cores. The SMM has 128 CUDA cores compared to 192 CUDA cores on the SMX.

The GM107 GPC ‘Graphics Processing Cluster’ consists of five of these streaming multiprocessors which are connected to a Raster Engine. Each SMM consists of Polymorph Engine 2.0 which includes the Vertex Fetch, Tessellator, Viewport transform, attribute setup and stream output. Each SMM has 8 texture mapping units which equates to 40 on the whole chip and 16 ROPs while connected to two 64-bit memory controllers.

There’s also a handful of cache on Maxwell which is 2 MB in total compared to 256 KB on Kepler GK107 which reduces GPU queuing. IPCC has been increased on the GPU core and balancing improvement to workload has been done. NVIDIA has improved their H.264 encoding and decoding with NVENC on Maxwell so that would result in better performance in ShadowPlay technology and improved sleep states have been implemented to reduce power input while the GPU is running idle. All of this is packed in a die which measures 148mm2  which means transistor density has been upped by 15% on the same 28nm process design.

  • NVIDIA GM200.GM210 (Maxwell Architecture, High-Performance for Telsa/Quadro Arrives later for Cosnumers, Successor of GK110)
  • NVIDIA GM204 (Maxwell Architecture, High-End Consumer, Successor of GK104, First GeForce 800 Series Products likely to feature)
  • NVIDIA GM206 (Maxwell Architecture, Performance Minded, Successor of GK206, Mid-Range GeForce 800 Series products to feature)
  • NVIDIA GM107/207 (Maxwell Architecture, Entry Level, Successor of GK107, Entry Level GeForce 800/700 Series To feature, Already introduced on GTX 750 Ti / GTX 750)

Do tell us if we missed out anything in the Disqus comments section below!

NVIDIA GeForce GTX 880 Specifications (Speculated):

GeForce GTX 480 GeForce GTX 580 GeForce GTX 680 GeForce GTX 780 GeForce GTX 780 Ti GeForce GTX 880
Codename GF100 GF110 GK104 GK110 GK110 GM204
Process 40nm 40nm 28nm 28nm 28nm 28nm
GPU Core Fermi Fermi Kepler Kepler Kepler Maxwell
Memory 1.5 GB 1.5 GB 2 GB 3 GB 3 GB / 6 GB 4-8 GB
Memory Bus 384-Bit 384-Bit 256-bit 384-Bit 384-Bit 256-bit
Launch March 26th 2010 November 09 2010 March 22nd 2012 May 23rd 2013 December 2013 September 2014
Price $499 US $499 US $499 US $499 US $699 US $549 US

 

NVIDIA GeForce GTX 880 vs GeForce GTX 680 ES PCB:

NVIDIA GeForce GTX 880 PCB Leaked Shots:

NVIDIA GeForce GTX 880 ES Board_2 NVIDIA GeForce GTX 880 ES Board_SK HynixNVIDIA GeForce GTX 880 16 GB MemoryNVIDIA GeForce GTX 880 ES Board_Power

Share Tweet Submit