Wonderful review as always! Would you consider including deep learning benchmarks in your future videos as that would be a massive deciding factor for people like me? Thanks!
Thanks for the feedback. Some deep learning benchmarks are definitely on our to-do list, we've had a play around with things like Tensorflow and Pytorch before so we'll probably use them so we can compare with old GPUs too. - JG
Very capable but expensive card. For non-cacheable algorithms like fluid simulations there is much less of a performance difference to the A6000, as raw VRAM bandwidth is only slightly faster. It's particularly interesting that NVLink was removed on Ada.
It's could be worth taking a look at this in-depth article from our friends at AEC Magazine, in which they tested the RTX 6000 Ada versus the RTX A6000. - JG
Would've been great to see the RTX 4090 in the lineup. I know it's not a "workstation" class card, but it can do all the same things the workstation graphics cards can do, and in most cases, can do them faster. For those of us who don't need a GPU on and working around the clock, in most cases the 76% increase in price brings almost no advantage if any. For those such as myself doing 3D animation and content creation, 48GB VRAM is always welcome, but not necessarily at the cost of slower compute performance, especially when considering that slower compute performance costs $5,200 more. I just don't get Nvidia's thinking. My organization could afford a 6000 Ada, but the added 76% price hike and slower compute performance just makes no logical sense. It's like finding a car that gets 100mpg but can only go 40mph. What's the point?
Thanks for feedback. Re your question about consumer versus workstation GPUs, the latter have always had lower compute performance. This is because they run at lower clock speeds, 1. to reduce the TDP so that 2-slot 'standard' coolers can be used versus the exotic 3+ slot coolers that most consumer GPUs use and 2. to improve reliability and longevity. We hear you on price, which is why as a workstation manufacturer we do offer both types of cards in our systems, so customers can decide what's most important to them. - JG
To keep the cars as an example, this is like complaining that your sports car is faster than a more expensive truck, and therefore it doesn't make sense to buy trucks at all.
@Frank Scheufens, That's a good analogy, gaming and workstation GPUs *can* both be used for compute, but they don't (by design) have the same characteristics. - JG
@@frankscheufens2992 you're a sage. The wisdom with which you explain is laudable. For people that don't need the truck due to the speed limitations and the bulky nature of the design, forgetting the reliability build can go for the fast 4090 and forget about longevity. I hope the 4090ti will be in the making
I'm thinking real hard about my options… My Dell T7910 can't fit even the smallest 4090 without a side cover mod. But more important - I don't want to throw away the benefits of this platform that has top end CPUs and 40 TB of NVMe and disk-RAID storage, ECC DDR4, and a 1300W Platinum PSW. Don't forget the 40xx video cards don't use ECC RAM. The RTX 6000 does. That's a BIG deal for the type of processing that I'm targeting, which is long term management of personal information, AI/ML, and maybe video rendering. I think many people would be surprised at how often bits are corrupted during normal processing by a number of "natural" factors and inconsistencies in silicon. Just do a search on studies that document how often RAM has errors, which leads to permanently storing that corruption. Or the problem of "bit rot" - which is degradation after storing data (why I like ZFS storage). I'm very interested in what others are thinking who are in my (somewhat legacy) boat... Pluses of RTX 6000 Ada: - Small size (fits my case) - Lower power consumption - Greater memory capacity - ECC memory - More processing units of the type I need (Tensor Cores, RT Cores, ROPs, TMUs, Shading Units) - Optimized by size and cooling to add more units in the system Minuses: - Cost! - Cost! - Cost! Do I want to pay $4K for a high-end GPU? Of course not! I have to weigh the cost and benefits. I'm not made of money. I have a lot of other uses for that cash. But I've always been willing to invest in my future if the reason is compelling. Would I play games on it? Hell NO! If gaming was my thing - I'd by a reasonable case, a MOBO that supports PCIe 4.0, probably an RTX 4070 and a high clock rate 6-8 core processor, 32GB DDR5, a 2TB PCIe 4.0 boot NVMe, plus a RAID 1 using 8TB WD Black or Red spinners for data, Thunderbolt 3.0 for a backup bus... and a high refresh rate 2K curved monitor. That would get the job done for the cost of ONE NVIDIA RTX 6000 Ada! Phew. That puts the cost/benefit analysis into perspective. But what do I give up with a gamer GPU? Robustness and total system capacity. That's the bottom line. The Dell T7910 has years of good service. Not everything I want, but will get the job done. Dual Xeon E5-2690v3's, 128 GB ECC RAM and several PCIe NVMe SSD's. I can upgrade the CPU's to v4, Plenty of room to upgrade storage to over 100TB and RAM to 512GB - possibly 1TB. I'd rather not constrain expansion of PCIe devices by having a 3.5x wide video card (!) because I still have room to add an NVMe RAID card that gets 10Gbs for a boot or cache drive. I don't want to add another 150W of power consumption, or add heat to the case with a card that doesn't push it out the back. That stresses the other components.
Wonderful review as always! Would you consider including deep learning benchmarks in your future videos as that would be a massive deciding factor for people like me? Thanks!
Thanks for the feedback. Some deep learning benchmarks are definitely on our to-do list, we've had a play around with things like Tensorflow and Pytorch before so we'll probably use them so we can compare with old GPUs too. - JG
Very capable but expensive card. For non-cacheable algorithms like fluid simulations there is much less of a performance difference to the A6000, as raw VRAM bandwidth is only slightly faster.
It's particularly interesting that NVLink was removed on Ada.
It's could be worth taking a look at this in-depth article from our friends at AEC Magazine, in which they tested the RTX 6000 Ada versus the RTX A6000. - JG
why does this channel only have 555 subscribers? well now it is 556 so i guess good luck
Would've been great to see the RTX 4090 in the lineup. I know it's not a "workstation" class card, but it can do all the same things the workstation graphics cards can do, and in most cases, can do them faster. For those of us who don't need a GPU on and working around the clock, in most cases the 76% increase in price brings almost no advantage if any. For those such as myself doing 3D animation and content creation, 48GB VRAM is always welcome, but not necessarily at the cost of slower compute performance, especially when considering that slower compute performance costs $5,200 more. I just don't get Nvidia's thinking. My organization could afford a 6000 Ada, but the added 76% price hike and slower compute performance just makes no logical sense. It's like finding a car that gets 100mpg but can only go 40mph. What's the point?
Thanks for feedback. Re your question about consumer versus workstation GPUs, the latter have always had lower compute performance. This is because they run at lower clock speeds, 1. to reduce the TDP so that 2-slot 'standard' coolers can be used versus the exotic 3+ slot coolers that most consumer GPUs use and 2. to improve reliability and longevity.
We hear you on price, which is why as a workstation manufacturer we do offer both types of cards in our systems, so customers can decide what's most important to them. - JG
To keep the cars as an example, this is like complaining that your sports car is faster than a more expensive truck, and therefore it doesn't make sense to buy trucks at all.
@Frank Scheufens, That's a good analogy, gaming and workstation GPUs *can* both be used for compute, but they don't (by design) have the same characteristics. - JG
It's a more stable card with more VRAM, also the slower compute is because of the low profile, they're working with a lower TGP
@@frankscheufens2992 you're a sage. The wisdom with which you explain is laudable.
For people that don't need the truck due to the speed limitations and the bulky nature of the design, forgetting the reliability build can go for the fast 4090 and forget about longevity.
I hope the 4090ti will be in the making
Can it run Minecraft in 4K?
Can play cyberpunk 2077
I would think I can, but not as good as a 30 or 40 series RTX GPU's or AMD's 7,000 series gaming GPUs in my option.
wtf this channel is amazing
hello girl
I'm thinking real hard about my options… My Dell T7910 can't fit even the smallest 4090 without a side cover mod. But more important - I don't want to throw away the benefits of this platform that has top end CPUs and 40 TB of NVMe and disk-RAID storage, ECC DDR4, and a 1300W Platinum PSW. Don't forget the 40xx video cards don't use ECC RAM. The RTX 6000 does. That's a BIG deal for the type of processing that I'm targeting, which is long term management of personal information, AI/ML, and maybe video rendering. I think many people would be surprised at how often bits are corrupted during normal processing by a number of "natural" factors and inconsistencies in silicon. Just do a search on studies that document how often RAM has errors, which leads to permanently storing that corruption. Or the problem of "bit rot" - which is degradation after storing data (why I like ZFS storage).
I'm very interested in what others are thinking who are in my (somewhat legacy) boat...
Pluses of RTX 6000 Ada:
- Small size (fits my case)
- Lower power consumption
- Greater memory capacity
- ECC memory
- More processing units of the type I need (Tensor Cores, RT Cores, ROPs, TMUs, Shading Units)
- Optimized by size and cooling to add more units in the system
Minuses:
- Cost!
- Cost!
- Cost!
Do I want to pay $4K for a high-end GPU? Of course not! I have to weigh the cost and benefits. I'm not made of money. I have a lot of other uses for that cash. But I've always been willing to invest in my future if the reason is compelling.
Would I play games on it? Hell NO! If gaming was my thing - I'd by a reasonable case, a MOBO that supports PCIe 4.0, probably an RTX 4070 and a high clock rate 6-8 core processor, 32GB DDR5, a 2TB PCIe 4.0 boot NVMe, plus a RAID 1 using 8TB WD Black or Red spinners for data, Thunderbolt 3.0 for a backup bus... and a high refresh rate 2K curved monitor. That would get the job done for the cost of ONE NVIDIA RTX 6000 Ada!
Phew. That puts the cost/benefit analysis into perspective. But what do I give up with a gamer GPU?
Robustness and total system capacity. That's the bottom line.
The Dell T7910 has years of good service. Not everything I want, but will get the job done. Dual Xeon E5-2690v3's, 128 GB ECC RAM and several PCIe NVMe SSD's. I can upgrade the CPU's to v4, Plenty of room to upgrade storage to over 100TB and RAM to 512GB - possibly 1TB. I'd rather not constrain expansion of PCIe devices by having a 3.5x wide video card (!) because I still have room to add an NVMe RAID card that gets 10Gbs for a boot or cache drive. I don't want to add another 150W of power consumption, or add heat to the case with a card that doesn't push it out the back. That stresses the other components.
There's my Ada😮❤
what's the physical weight of this card?
Seems like it has more copper inside of it than most consumer cards do.
yeah but how many fps does it get in cs:go
Great video!
Thanks, glad you enjoyed it. - JG
Any idea about the launch of other RTX Ada GPUs? 4500 or maybe 4000. I really want to build a workstation but my pocket doesnt allow for 6000 Ada.
NVIDIA's GTC conference is in about a week's time so there might be something announced then, but no confirmed news yet I'm afraid. - JG
@@scanprographics5094 auryt. Thanks !!!
I5⏰4460 🛌🏻