[Rumor, Tweaktown] AMD to launch next-gen Navi graphics cards at E3

[beginner99]

1/2 node advantage (12 to 7 isn't a full node)

12nm is just a fancy name for 16nm optimized for NV (clocks, reticle limit). Don't get lost in marketing terms.

Perf:Watt has nothing to do with actual performance though

No but it pretty much sets a limit how high you can go with performance. All things equal including price, which card to you buy? The one that uses less or more power? If you price your product so close to perfomrance/$ than competition other factors start to matter. For me it isn't the power that matters but card size, fan noise and most importantly waste heat. It matters to me due to experience because running a 290x in summer without AC sucks. Even in deep winter it's obvious the room is much warmer after 1hr of gaming.

 

[Glo.]

Show me precedent for any company in CPU, GPU space intentionally showing lower numbers at one of the common tech conventions.

So you do not have any proof for that, AMD has shown best case scenario, and this is only your assumption.

 

[Glo.]

According to ElChapuzasInformatico's review of Ryzen 5 3600, it appears that 7 nm process is not as good as everybody has hoped for. 3600 is consuming just 10W's less than 2600.

Who knows, maybe my words about N7 process node are correct, don't ya think?

 

[DarthKyrie]

According to ElChapuzasInformatico's review of Ryzen 5 3600, it appears that 7 nm process is not as good as everybody has hoped for. 3600 is consuming just 10W's less than 2600.

Who knows, maybe my words about N7 process node are correct, don't ya think?

Remember that the N7 AMD is using is the HPC variant that they aided TSMC in the design of, it is going to be a power hog for the node size.

 

[DrMrLordX]

But the mention of Navi 21 is new. This may be the replacement for the Vega VII for use in data centers?

I was under the impression that Navi/RDNA is consumer-only while iterations of GCN (as we know it) will continue in the datacenter. If Navi winds up in datacenters, I'd be surprised.

 

[DarthKyrie]

I was under the impression that Navi/RDNA is consumer-only while iterations of GCN (as we know it) will continue in the datacenter. If Navi winds up in datacenters, I'd be surprised.

I think AMD will have an RDNA GPU for compute at some point but that is still a few years away.

 

[Ajay]

I was under the impression that Navi/RDNA is consumer-only while iterations of GCN (as we know it) will continue in the datacenter. If Navi winds up in datacenters, I'd be surprised.

Yeah, there could be a workstation variant of Navi (it’s really about the drivers). But, data center, no RAS, no ECC, etc..

 

[Stuka87]

I was under the impression that Navi/RDNA is consumer-only while iterations of GCN (as we know it) will continue in the datacenter. If Navi winds up in datacenters, I'd be surprised.

This is what I was under the impression of as well. Which is why I worded it as a question. It could be just workstation like the above posts mention. But it could be that Navi 21 doesn't ship till late next year, which would work out fine as a Vega replacement on a time window basis. AMD stated that GCN would live in for compute, but they didn't say navi would never replace it, just that it wont near term.

 

[crisium]

Blaming nodes for power consumption again? I remember it well, blaming GF 14nm for Polaris' shortcomings back in 2016. Then we get 12nm Polaris and it still can't compete with Nvidia efficiency, but I guess we'll just blame the node again for Navi.

 

[Guru]

According to ElChapuzasInformatico's review of Ryzen 5 3600, it appears that 7 nm process is not as good as everybody has hoped for. 3600 is consuming just 10W's less than 2600.

Who knows, maybe my words about N7 process node are correct, don't ya think?

I'd wait for more reliable reviewers, especially any claim of a review this early is probably garbage. We know that review chips usually arrive about 7-10 days before official release/reveal day, sometimes even just 3-4 days before reveal. So having a review chip almost 3 weeks before release date to me seems like its 100% fake clickbait news.

There is no way AMD are able to get 16 cores 32 threads at 4.7GHz at 105W if 7nm wasn't MUCH better than their 16nm process. Its just ridiculous that anyone would suggest that 7nm is barely better than 16nm for AMD, that is absurd. Just the size difference, density, performance, etc.... that AMD can get out of 7nm vs 16/12nm is huge.

Again its the Intel sponsored fake news, because they have no competition to AMD, so lets activate all our old pals, all our paid promoters and have them spread fake news.

I'd wait for real reviews like Anandtech, Guru3d, computerhardware.de, few others that are still reliable.

 

[Glo.]

I'd wait for more reliable reviewers, especially any claim of a review this early is probably garbage. We know that review chips usually arrive about 7-10 days before official release/reveal day, sometimes even just 3-4 days before reveal. So having a review chip almost 3 weeks before release date to me seems like its 100% fake clickbait news.

There is no way AMD are able to get 16 cores 32 threads at 4.7GHz at 105W if 7nm wasn't MUCH better than their 16nm process. Its just ridiculous that anyone would suggest that 7nm is barely better than 16nm for AMD, that is absurd. Just the size difference, density, performance, etc.... that AMD can get out of 7nm vs 16/12nm is huge.

Again its the Intel sponsored fake news, because they have no competition to AMD, so lets activate all our old pals, all our paid promoters and have them spread fake news.

I'd wait for real reviews like Anandtech, Guru3d, computerhardware.de, few others that are still reliable.

105W for 16C CPU is just the Cooler TDP rating, not the actual power draw. If 6C/12T CPU is at least 65W(more like 70-75W), don't expect that 16C part will use less than 125W of power. The process is not that better than 16/14 nm in terms of power consumption at maximum clocks.

Your posts's part about Intel payed review is pure BS. ElChapuzas, most likely, bought Engineering Sample, as they did when first Zen 1 CPU review appeared. They were the first to publish LEGIT Zen 1 review. The thing is, they were done on wrong BIOSes, and with bad memory. Here we have bad memory scores, which suggests that BIOS is not mature enough. But it is 100% legit review.

 

[Glo.]

Blaming nodes for power consumption again? I remember it well, blaming GF 14nm for Polaris' shortcomings back in 2016. Then we get 12nm Polaris and it still can't compete with Nvidia efficiency, but I guess we'll just blame the node again for Navi.

Why don't we wait for Navi reviews and then lose our minds about this GPU architecture, eh?

 

[Guru]

105W for 16C CPU is just the Cooler TDP rating, not the actual power draw. If 6C/12T CPU is at least 65W(more like 70-75W), don't expect that 16C part will use less than 125W of power. The process is not that better than 16/14 nm in terms of power consumption at maximum clocks.

Judging by AMD's 1000 and 2000 series CPU's, their processors operate WITHIN the advertised power envelope assuming everything i8s standard. You can read Anandtech review right now or better yet tomshardware review and find out that their processors operate within the power envelope under normal circumstances.

If you have significant cooling, then XFR kicks in and boosts additional cores by up to 200mhz, which can increase the wattage beyond advertised levels, but for the most part AMD processors have operated within or close to their advertised power envelopes, unlike Intel who've been cheating by basing their TDP under base frequency.

Again I'm expecting much better boost algorithm, boosting more on fewer cores to push the processor in lightly threaded applications, I'm expecting higher clocks in general and just overall good 15% performance improvement over Ryzen 2000 across the board at slightly less power consumption including with higher clocks. To me that is a huge success.

 

[Glo.]

Judging by AMD's 1000 and 2000 series CPU's, their processors operate WITHIN the advertised power envelope assuming everything i8s standard. You can read Anandtech review right now or better yet tomshardware review and find out that their processors operate within the power envelope under normal circumstances.

If you have significant cooling, then XFR kicks in and boosts additional cores by up to 200mhz, which can increase the wattage beyond advertised levels, but for the most part AMD processors have operated within or close to their advertised power envelopes, unlike Intel who've been cheating by basing their TDP under base frequency.

Again I'm expecting much better boost algorithm, boosting more on fewer cores to push the processor in lightly threaded applications, I'm expecting higher clocks in general and just overall good 15% performance improvement over Ryzen 2000 across the board at slightly less power consumption including with higher clocks. To me that is a huge success.

No, they are not operating at advertised power envelope, at least when it comes for power draw.

https://www.computerbase.de/2018-06/intel-core-i7-8700t-i5-8500t-cpu-test-coffee-lake/3/
You can compare Core i7 8700T locked to 35W, which resulted in 63W power draw for whole platform. Ryzen 5 2400G which is the closest of all AMD CPUs to 65W TDP rating, in power draw consumed 67W more for whole platform.

2600 platform is using 146W of power in this particular test. 83W more. You think that MoBo and RAM would consume more power, than the CPU, alone? I don't think so. Its safe to assume that Ryzen 5 2600 is in reality not 65W power draw CPU, but more like 85-90W CPU.

 

[crisium]

Why don't we wait for Navi reviews and then lose our minds about this GPU architecture, eh?

No.

It's very fair to react based on AMD's publicly revealed information.

 

[Stuka87]

No, they are not operating at advertised power envelope, at least when it comes for power draw.

https://www.computerbase.de/2018-06/intel-core-i7-8700t-i5-8500t-cpu-test-coffee-lake/3/
You can compare Core i7 8700T locked to 35W, which resulted in 63W power draw for whole platform. Ryzen 5 2400G which is the closest of all AMD CPUs to 65W TDP rating, in power draw consumed 67W more for whole platform.

2600 platform is using 146W of power in this particular test. 83W more. You think that MoBo and RAM would consume more power, than the CPU, alone? I don't think so. Its safe to assume that Ryzen 5 2600 is in reality not 65W TDP CPU, but more like 85-90W CPU.

When you say "whole platform" are you talking motherboard and memory included? If so, than thats an incorrect power measurement. Motherboard and memory power draw are out of the hands of the CPU maker. They list what the peak power rating of the chip is. The motherboards then state what TDP's their power system will support. Any power used over that by the board and memory are above and beyond what the CPU itself uses and do not invalidate what intel/AMD state for TDP.

 

[ubern00b]

No, they are not operating at advertised power envelope, at least when it comes for power draw.

https://www.computerbase.de/2018-06/intel-core-i7-8700t-i5-8500t-cpu-test-coffee-lake/3/
You can compare Core i7 8700T locked to 35W, which resulted in 63W power draw for whole platform. Ryzen 5 2400G which is the closest of all AMD CPUs to 65W TDP rating, in power draw consumed 67W more for whole platform.

2600 platform is using 146W of power in this particular test. 83W more. You think that MoBo and RAM would consume more power, than the CPU, alone? I don't think so. Its safe to assume that Ryzen 5 2600 is in reality not 65W TDP CPU, but more like 85-90W CPU.

TDP is not the same thing as power consumption though..

 

[Bouowmx]

Before we dive into the deep end, I want to say: try not to stray into CPU power limits discussion (again?). Remember, this thread is about Navi

 

[railven]

I feel like one thing that probably isn't covered as much in regards to the stagnation is consoles. Gen 8th launched with some horrendous choices causing them to basically launch with obsolete hardware. There wasn't much room to push progress. Why we saw NV push more of their technologies, to promote need for stronger hardware.

I hope these next consoles don't launch as week as Gen 8. The console section as a whole is also seeing a decline with the rapid growth of mobile. Probably the best decision Nintendo did was making their console a hybrid. It's close to outselling Xbox in 1/3 of the time.

Stronger consoles -> more advance games -> more use of processing power

 

[Glo.]

No.

It's very fair to react based on AMD's publicly revealed information.

And what you can draw out of AMD's publicly disclosed information about Navi's Power draw, hm?

 

[Glo.]

TDP is not the same thing as power consumption though..

Which is exactly what I was alluding to . And that AMD CPUs are using more power than their TDP ratings would suggest.

 

[crisium]

And what you can draw out of AMD's publicly disclosed information about Navi's Power draw, hm?

AMD stated it's 14% faster than Vega 64 for 23% less power.

Hm? Was that supposed to be a got ya? It's not like we have Vega 64 numbers out there for 2 years to extrapolate off this or anything.

 

[ubern00b]

Which is exactly what I was alluding to . And that AMD CPUs are using more power than their TDP ratings would suggest.

Their power draw has nothing to do with their TDP, the same is true for Intel, so it makes no difference. if it was 65w TDP CPU with 65W power consumption then it's just a heater as it's converting all electrical input into heat output, as it stands, you can take 150w of energy and only need 65w of thermal dissipation to remove the heat that's been converted, seems some people can't make the differentation

 

[Glo.]

Their power draw has nothing to do with their TDP, the same is true for Intel, so it makes no difference. if it was 65w TDP CPU with 65W power consumption then it's just a heater as it's converting all electrical input into heat output, as it stands, you can take 150w of energy and only need 65w of thermal dissipation to remove the heat that's been converted, seems some people can't make the differentation

For **** sake. Read the post to which I was replying in the first place. Then disagree with me. Also how can you disagree with me and then basically agree with me, but display it as being meaningless?

This is the post that I was replying to:

I'd wait for more reliable reviewers, especially any claim of a review this early is probably garbage. We know that review chips usually arrive about 7-10 days before official release/reveal day, sometimes even just 3-4 days before reveal. So having a review chip almost 3 weeks before release date to me seems like its 100% fake clickbait news.

There is no way AMD are able to get 16 cores 32 threads at 4.7GHz at 105W if 7nm wasn't MUCH better than their 16nm process. Its just ridiculous that anyone would suggest that 7nm is barely better than 16nm for AMD, that is absurd. Just the size difference, density, performance, etc.... that AMD can get out of 7nm vs 16/12nm is huge.

Again its the Intel sponsored fake news, because they have no competition to AMD, so lets activate all our old pals, all our paid promoters and have them spread fake news.

I'd wait for real reviews like Anandtech, Guru3d, computerhardware.de, few others that are still reliable.

 

[Glo.]

AMD stated it's 14% faster than Vega 64 for 23% less power.

Hm? Was that supposed to be a got ya? It's not like we have Vega 64 numbers out there for 2 years to extrapolate off this or anything.

Yes, but you have not read the caveats. 40 CU chip was 14% faster than 64CU Vega, and 40 CU Navi chip was using 23% less power THAN 40 CU VEGA chip.

What can you draw from this?