Is this all a dream?
https://blogs.nvidia.com/blog/2017/07/17/intel-amd-accelerated-data-center/
>>61427597
NVidia with the counter-counter shitpost.
>>61427597
>scientific understanding
>>61427597
if all you want to do is run a heap of gpus, amd have the better option now (all those fucking pcie lanes out of one socket)
amd can win even if everyone picks nvidia for gpus
kek
>>61427736
>NVidia helps AMD take over the datacenter
It's like pottery.
>buy EPYC
>shitload of PCIe lanes
>cheaper than the Intel alternative, that has less lanes
>use the saved money for buying more GPUs
>no point in falling for the AVX meme
It's a win-win situation for both AMD and Nvidia.
>>61427775
Don't you mean poetry?
>>61428404
Newfig.
>>61428153
Oy! Delet.
>>61427597
NVIDIA cares not about supplier of PCI-E lanes for their goy processing units.
But considering EPYC allows you to strap whole 6 goy processing units to one board, they'll shill it to no end despite JHH probably STILL being mad at AMD.
>>61428565
Why is Huang so mad at AMD?
>>61428613
AMD bought ATi. nvidia and AMD used to be best friends forever before that.
>>61427775
There's a reason for that. Intel's the main adversary of Nvidia in the field of big data. If nVidia can make AMD and Intel fight it over and reduce Intels' money, this is a win for them.
>>61428638
>this is a win for them
AMD having fuckton of datacenter money for their every funky R&D venture is nvidia's worst nightmare actually.
>>61428661
Yeah, but Nvidia tried to get the x86 license from Intel in a cross-licensing agreement. Intel said no, but instead paid Nvidia $1Bn in cash for access to some of their GPU IPs. Nvidia's been massively butthurt about that; and as a result, has invested heavily in cars, simulation, AI, etc.
They wanted the low-end/mobile market that AMD has access to, and Intel has access to; as their gaming margins wouldn't sustain them forever. All that said, AMD with datacenter revenue going into R&D is still not a threat to Nvidia, since their entire company is several times larger in terms of a monetary footprint than all of AMD & RTG combined.
They are willing to take the risk of a more positively funded AMD for CPU/GPU R&D than dealing with Intel's shenanigans. Intel is a massive jew, and Nvidia is also a massive jew; and there can only be one ultra jew in the CPU/GPU business.
Finally, a 1P EPYC chip gives 64 PCI-E lanes for GPUs. That translates into 4 GPUs at full 16x bandwidth or 8 GPUs at 8x bandwidth; either way, its a win for Nvidia--because anyone investing into EPYC will have more money left over for Nvidia hardware. Whereas, anyone investing into an equivalent Intel config, will have more upfront overhead before they order Nvidia hardware.
And in HPC/AI, where DNN is king, having more GPUs in a single rack is far more important than having very highly clocked CPU cores that have monstrous IPC, as that is not particularly important.
Nvidia gains nothing from Intel's long-term jewery, and gains everything, from AMD taking 20-30% of Intel's datacenter marketshare.
>>61428661
Its CPU and GPU teams are different. Profits gained on CPU side will be invested more into CPU. GPU side might see some additional RD money but nothing that would be considered "fuckton".
AMD and Intel fighting over datacenter money won't result in "fuckton" of money for AMD. Its a pretty tight race so while AMD might gain good shares there, its a tight spot.
That said, even if AMD doubles or triples their net income, it still wont be enough to combat either Nvidia or Intel's R/D or their income. Both Intel/Nvidia are that much larger. The real role AMD will play is to distract Intel and take bit off their shares. Hurting Intel will be in nVidia's favor, even if little.