-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
3950x coming September 30th (WCCFTech)
- Thread starter pgdeaner
- Start date
- Joined
- Mar 7, 2008
What a time we live in, when my first thought was "only" 24 cores? 
Wonder if AND will go for another "jebait" claim. Intel's next HEDT is based on Cascade Lake, which offers up to 28 cores per die on the Xeon side. We don't know if they will necessarily offer that many in HEDT. Anyway, AMD announces 24, Intel trumps with 28, AMD conveniently remembers it has a 32 core model too. Boom.
Think at this point AMD can do pretty much whatever they want product wise. Even if they don't want to encroach on 64 core Epycs, they could still upscale TR to 48 cores if they need/want to. This is on assumption Intel isn't going to use Cascade Lake AP for HEDT where 56 is possible.
Wonder if AND will go for another "jebait" claim. Intel's next HEDT is based on Cascade Lake, which offers up to 28 cores per die on the Xeon side. We don't know if they will necessarily offer that many in HEDT. Anyway, AMD announces 24, Intel trumps with 28, AMD conveniently remembers it has a 32 core model too. Boom.
Think at this point AMD can do pretty much whatever they want product wise. Even if they don't want to encroach on 64 core Epycs, they could still upscale TR to 48 cores if they need/want to. This is on assumption Intel isn't going to use Cascade Lake AP for HEDT where 56 is possible.
I guess that not long after the premiere we will see 32 and 64 cores too. What is more interesting for me is how will be PCI-E lane distribution and memory controller. I just wonder if they use more than 4 memory channels like in Epic or keep 4. If current X399 motherboards will support new chips then 4 channels will be max.
- Joined
- Aug 11, 2015
And still only 4 channels for memory, though 8 lanes to the PCH and I/O die. But not backward or forward compatible with X399. Have to move to the new TRX40 platform to use the new Threadripper 3 cpus.
That's a bit annoying as AMD said that will use current sockets at least in one more generation. It looks like AM4 is maybe the same but because of chipset and CPU (with all integrated stuff) changes, it requires to change the motherboard or something won't work as expected. In the case of TR, there is a new chipset and motherboards so all who invested in this platform have no way to upgrade other than to replace the most expensive components. Also, older stuff lost a lot on the value, and also because of AMD price drops so there is no way to get money back to cover at least half of the hardware (considering expected prices) ... not to mention how hard it will be to sell it.
Resources... or lack there of.
Looks like the Tom's article it boosted well. It just varies by sample it seems. If that is true, thats'........for lack of a better term, whack. If the box says x.x GHz, it should reach x.x GHz on a holy ton of them.
Looks like the Tom's article it boosted well. It just varies by sample it seems. If that is true, thats'........for lack of a better term, whack. If the box says x.x GHz, it should reach x.x GHz on a holy ton of them.
- Joined
- Mar 7, 2008
Personally I prefer written over video reviews as generally there is a lot of info to digest, and video isn't the best format for that unless you just want a casual high level summary.
The boost thing, is it still a thing? Without digging into exactly how each place tested, it is an "up to" speed that wont be reached by all workloads. We're now spreading the power budget over more cores so it may be ever more challenging to encounter that case.
The boost thing, is it still a thing? Without digging into exactly how each place tested, it is an "up to" speed that wont be reached by all workloads. We're now spreading the power budget over more cores so it may be ever more challenging to encounter that case.
If it's challenging, then perhaps the specification should be lowered? I don't know. I'm from the school of if it says it on the box, it should reach it. I don't like this 'up to' stuff from anyone (AMD GPU Boost clocks, anyone?). The difference between the two results on high-end boards and cooling is too much to me. That is a difference of what, 200 Mhz?
- Joined
- Mar 7, 2008
It is a challenging one. I don't like "up to" speeds either, as usually it is used by marketing to make something sound better than it is. Part of the problem as I see it is that there is a window where you can load the CPU up without hitting power, thermal or current limits, where you can get the maximum boost. Without more budget in those areas, adding more cores narrows that window. If we take it the other direction, we end up with Intel's TDP. The minimum clock you get at maximum operating temperature, which is achieved with a cooler working at that TDP. But that's too far the other direction as most of the time you will be somewhere above it. This is essentially another case where we're looking for a complicated description, that marketing wants to simplify to one or two numbers. If you can come up with a good solution to this, you're better than I am, and perhaps most of the industry also.
About the only way I can think of possibly getting something like this is an "average clock" rating running some mixed workloads, but there will always be arguments over the choice of that mix, and it will need to change over time, will depend on the rest of the system specifications... you might as well run a random number generator at that point.
About the only way I can think of possibly getting something like this is an "average clock" rating running some mixed workloads, but there will always be arguments over the choice of that mix, and it will need to change over time, will depend on the rest of the system specifications... you might as well run a random number generator at that point.
Or why not raise the tdp a little so it can sustain it? It doesn't have to get to 'intel tdp' levels, just go from 105W to 115W or whatever it takes.
I know the performance delta isn't much, but still. If it says x.x on the box, it should reach that on way more samples than not.
I know the performance delta isn't much, but still. If it says x.x on the box, it should reach that on way more samples than not.
Alaric
New Member
- Joined
- Dec 4, 2011
- Location
- Satan's Colon, US
Or, Linus' take on it if that's more your style. He seems quite impressed with the numbers generated, including power draw.
He seems quite impressed with the numbers generated, including power draw.
Alaric
New Member
- Joined
- Dec 4, 2011
- Location
- Satan's Colon, US
He covers the high points fairly well in that video, at least the points that most end users will care about.
Similar threads
