Quarter 1 Fiscal Year 2018 earnings. NVIDIA's fiscal year is about 11 months ahead of the calendar year. Their Fiscal Year 2017 ended on January 29, 2017.
In general you can't directly match up segments of different companies. How they split up their revenue into reporting segments and what they name those segments is up to them. In any case, as far as I know, AMD reports XBOX and PS4 revenue under their 'Embedded and Semi-custom' segment.
NVIDIA's 'OEM & IP' segment includes payments to NVIDIA from Intel stemming from a lawsuit settlement some years back (the 'IP'). Since the settlement, NVIDIA has reported $66M each quarter from those payments, but they are now ending. For the reported quarter, NVIDIA recognized the last $44M (IIRC) in IP payment instead of the full $66M, which results in the reported lower 'OEM & IP' revenue. The type of revenue placed in the 'OEM' portion of that segment is also something that NVIDIA has been de-emphasizing in their business for the last few years because it has lower margins than most of the rest of their revenue.
I just looked it up on their website. The correct amount of revenue from the Intel patent licensing agreement that NVIDIA recognized for this reported quarter is $43M, not $44M.
No. The CUDA cores are very good for highly parallel (matrix/vector) calculations. The Titan Xp costs far less than a Tesla card and is designed to go in a workstation instead of a server. The Machine Learning crowd are crazy about these cards.
There is nothing the Titan Xp offers over a GTX 1080Ti for those applications. It's just a gaming card being given a spin by NVIDIA to justify its existence and price.
You didn't hear the same marketing for the Titan X(Pascal), did ya?
They didn't need to do any marketing spin on the Titan X(Pascal) because it's nearly twice as fast as the Maxwell version in Single-precision and was the first Titan card with FP16, which is actually better for machine learning. Whereas the Titan Xp is just the fully enabled GP102 die, which isn't that much better than the Titan X Pascal.
As for the Titan Xp vs the 1080Ti, any gaming enthusiast who buys it is just throwing away their money. I can definitely see researchers upgrading from the Maxwell Titan X to the Titan Xp. The extra 60GB/s of memory bandwidth and 1GB of memory (compared to the 1080 Ti) can make a fair difference when working with larger data sets.
1080Ti has support for FP16 as well. The extra 9% memory and 60GB/s bandwidth would not ultimately matter that much.
It is a marketing spin to lure gullible customers. There are exactly zero reasons to buy Titan X cards since GK110, unless you want to pay the NVIDIA tax for early adoption and the absolute best performance..
Titan X Maxwell had FP16 support which NVIDIA felt that it cannibalized sales of the M40. Since Pascal there is very limited support for FP16 on Titan XP and Titan Xp.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
16 Comments
Back to Article
auralcircuitry - Tuesday, May 9, 2017 - link
...2018 earnings?bcronce - Tuesday, May 9, 2017 - link
Nvidia powered AI nural-nets have deep-learned and predicted the future.Yojimbo - Tuesday, May 9, 2017 - link
Quarter 1 Fiscal Year 2018 earnings. NVIDIA's fiscal year is about 11 months ahead of the calendar year. Their Fiscal Year 2017 ended on January 29, 2017.webdoctors - Tuesday, May 9, 2017 - link
money money money!Any info on why OEM & IP section has gone down 10%? Isn't that AMD's cash cow right now with XBOX and PS4?
Brett Howse - Tuesday, May 9, 2017 - link
They used to receive $66 million per quarter from Intel for a settlement, and I believe that has now ended. http://www.anandtech.com/show/4122/intel-settles-w...I thought this was already done but I believe it actually is now. Intel and NVIDIA report it differently though.
Yojimbo - Tuesday, May 9, 2017 - link
In general you can't directly match up segments of different companies. How they split up their revenue into reporting segments and what they name those segments is up to them. In any case, as far as I know, AMD reports XBOX and PS4 revenue under their 'Embedded and Semi-custom' segment.NVIDIA's 'OEM & IP' segment includes payments to NVIDIA from Intel stemming from a lawsuit settlement some years back (the 'IP'). Since the settlement, NVIDIA has reported $66M each quarter from those payments, but they are now ending. For the reported quarter, NVIDIA recognized the last $44M (IIRC) in IP payment instead of the full $66M, which results in the reported lower 'OEM & IP' revenue. The type of revenue placed in the 'OEM' portion of that segment is also something that NVIDIA has been de-emphasizing in their business for the last few years because it has lower margins than most of the rest of their revenue.
Yojimbo - Tuesday, May 9, 2017 - link
BTW, NVIDIA's Nintendo Switch revenue is reported under 'Gaming'.Yojimbo - Tuesday, May 9, 2017 - link
I just looked it up on their website. The correct amount of revenue from the Intel patent licensing agreement that NVIDIA recognized for this reported quarter is $43M, not $44M.StrangerGuy - Wednesday, May 10, 2017 - link
Who cares about the breakdowns? When it's obvious who has a GPU business that is run as a for-profit, and who has one that is run as a charity.Tamz_msc - Wednesday, May 10, 2017 - link
"Announced TITAN Xp for enthusiasts and researchers requiring extreme performance."Yeah researchers want to play AAA games at 4K 60FPS.
SaberKOG91 - Wednesday, May 10, 2017 - link
No. The CUDA cores are very good for highly parallel (matrix/vector) calculations. The Titan Xp costs far less than a Tesla card and is designed to go in a workstation instead of a server. The Machine Learning crowd are crazy about these cards.Tamz_msc - Wednesday, May 10, 2017 - link
There is nothing the Titan Xp offers over a GTX 1080Ti for those applications. It's just a gaming card being given a spin by NVIDIA to justify its existence and price.You didn't hear the same marketing for the Titan X(Pascal), did ya?
SaberKOG91 - Wednesday, May 10, 2017 - link
They didn't need to do any marketing spin on the Titan X(Pascal) because it's nearly twice as fast as the Maxwell version in Single-precision and was the first Titan card with FP16, which is actually better for machine learning. Whereas the Titan Xp is just the fully enabled GP102 die, which isn't that much better than the Titan X Pascal.As for the Titan Xp vs the 1080Ti, any gaming enthusiast who buys it is just throwing away their money. I can definitely see researchers upgrading from the Maxwell Titan X to the Titan Xp. The extra 60GB/s of memory bandwidth and 1GB of memory (compared to the 1080 Ti) can make a fair difference when working with larger data sets.
Tamz_msc - Thursday, May 11, 2017 - link
1080Ti has support for FP16 as well. The extra 9% memory and 60GB/s bandwidth would not ultimately matter that much.It is a marketing spin to lure gullible customers. There are exactly zero reasons to buy Titan X cards since GK110, unless you want to pay the NVIDIA tax for early adoption and the absolute best performance..
Tamz_msc - Thursday, May 11, 2017 - link
Titan X Maxwell had FP16 support which NVIDIA felt that it cannibalized sales of the M40. Since Pascal there is very limited support for FP16 on Titan XP and Titan Xp.helvete - Monday, July 31, 2017 - link
TITAN Xp cannot run AAA games at 4K 60FPS..