So lets say an i5 (7600K) vs i7 (7700K). Both at same speed. Just what is the difference. I remember when I bought my i5 I was told by many that for gaming don't get an i7, get the i5. But what is the case now? Me personally I do quite a bit of video editing and graphical things now, so I'm just wondering how much of a difference is 8 threads over 4 threads? Is it twice as fast in rendering? or only 5% extra speed? Just how does it work out. Thanks.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Hyperthreading vs a physical core.
- Thread starter beany_bot
- Start date
Depends hugely application to application. Typically the advantage of HT is more in the 25-30% region but some applications can see 50+% or even as high as 95% from the additional HT units. Personally I consider it worth it for gaming especially if you are running something beyond 60Hz with V-Sync as even when the performance isn't necessarily much different it is often smoother these days as a lot of games use 1-2 main threads and then spawn a lot of worker threads which seems to result in less stutter, etc. if they are kept away from the busier cores.
Yeah I'm excited about upgrading from my 4690K to either a hyperthreaded 6 core i7 or 8 core Ryzen 7.
The odd thing is... I always thought my GPU would be the first thing to upgrade. (As I haven't really upgraded in ages going from 390x -> 480 -> 1060. (Basically all perform the same). but god damn the cost of high end GPU's now is just stomach turning. Plus I'm gaming much less these days. Doing more content creation for a few groups I manage. so oddly the CPU might be first to change.
The odd thing is... I always thought my GPU would be the first thing to upgrade. (As I haven't really upgraded in ages going from 390x -> 480 -> 1060. (Basically all perform the same). but god damn the cost of high end GPU's now is just stomach turning. Plus I'm gaming much less these days. Doing more content creation for a few groups I manage. so oddly the CPU might be first to change.
This is enormous speculation but do you think for Intels 9th gen. They will up to 8 cores 16 threads to match Ryzen? or will we be sticking with 6C/12T for a while for mainstream i7's? I guess it depends what AMD do next. If they start knocking on Intels door in terms of IPC and clock speed then Intel will have to match them in core count.
I'm going to say it entirely depends on AMD now. As it stand Intel don't need to do 8 cores because they have the clock and IPC win. So most gamers will still be buying Intel. However if AMD get close to Intel clock and IPC then Intel will simply have to match them core for core. Adding cores to a CPU is a real cost per CPU that Intel don't want to incur. That said, I really don't think Ryzen will be able to hit that Intel calculation speed. I think Intel know that. So are just doing "enough" to stop gamers buying Ryzen (which that have been doing of late).
Especially as the fabrication die size has been the same since Skylake. I hope Cannonlake will introduce the 8 cores 16 threads on 10nm.
I agree largely, but I think that there is still an attraction in Ryzen to the less deep pockets between the i5 and the 1600 or less. The IPC is a little less, maybe 5-15%, this translates to a very few frames per sec to most people and it does have a marketable hook for OEM's and system builders.
There are still many 7th gen Intel out there to compete with.
Doubt it.
Man of Honour
- Joined
- 12 Jul 2005
- Posts
- 21,492
- Location
- Aberlour, NE Scotland
It look's like Intel will be launching 8 core, 16 thread cpu's next year but they will require yet another chipset (Z390) and are not backwards compatible with Z370. Article here.
That's still Coffee Lake, not Cannonlake (if the slide is accurate). Interesting that it'll take them a year to slap on another two cores - I guess not being able to glue them together makes it more difficult?
I predict:
Q4 2018: 8c/16t Coffee Lake
Q4 2019: Icelake
Interesting that it'll take them a year to slap on another two cores - I guess not being able to glue them together makes it more difficult?
To be fair, AMD's fabric/glue seems to have been utter genius. The ability to pump out just one type of silicon then stitch it together has let them scale up to TR and Epic relatively cheaply. Intel have missed a trick with their monolithic dies, methinks...
To be fair, AMD's fabric/glue seems to have been utter genius. The ability to pump out just one type of silicon then stitch it together has let them scale up to TR and Epic relatively cheaply. Intel have missed a trick with their monolithic dies, methinks...
I'm waiting on the Samsung Evostik.................. I'll get my coat.
To be fair, AMD's fabric/glue seems to have been utter genius. The ability to pump out just one type of silicon then stitch it together has let them scale up to TR and Epic relatively cheaply. Intel have missed a trick with their monolithic dies, methinks...
It's not genius, it's cheap. It introduces a load of communication overhead between the dies. Notable that Intel did two dies stuck together in the olden days, at least as far back as the q6600.
Thats mad, it's a lossless communication system. quite literally "perfect". From what I have seen.
Good question. Traditionally it's changed when they go for significant architecture changes, i.e. Sandy Bridge & Ivy Bridge, followed by Haswell & Broadwell, followed by Skylake, Kaby Lake, Coffee Lake & Cannonlake, which all makes sense...but then there's Icelake, which is meant to be a significant architecture change. So...who knows?
That it's tied to memory speed does seem to be a drawback, but AFAIK the scaling between 1 CCX, 2, or 4, is almost perfect. Cross-core communication doesn't seem to suffer any when stepping up to TR or Epyc. Unless I'm misinformed
That it's tied to memory speed does seem to be a drawback, but AFAIK the scaling between 1 CCX, 2, or 4, is almost perfect. Cross-core communication doesn't seem to suffer any when stepping up to TR or Epyc. Unless I'm misinformed
The latency in epyc is higher than the equivalent xeon setup between 2 cpu's. They need to untie it or find a faster IF.