What I found about 5775c EDRAM's impact on gaming performance.

[blobster21]

Can you show me a picture of the EDRAM control in the Z97X gaming bios ?

 

[R0H1T]

It's accessible for cpu and igpu both. Not much to be found as for its function, but from what I understand it's like a low latency buffer between L3 and DRAM. Whatever its purpose, the benefits for cpu limited scenarios are huge.



was gonna pick up wildlands when the price drops, it did drop but now it jumped back so no go, not gonna pay that much for what we in Poland call "a reheated pork chop".
That said, many new games need ridiculous CPU resources. WatchDogs2 is one I like benchmarking, I found certain parts of Dishonored 2 to be a CPU hog as well. Battlefield 1 is an obvious example,Deus Ex Mankind Divided also shows huge demand for cpu resources.


testing The Witcher 3 now, Novigrad (cpu heavy), 1080p, low

CPU 3.7GHz L4 disabled - 130 fps
CPU 4.2GHz L4 disabled - 141 fps

edram 1.8GHz tests coming in a while.

CPU 3.7 L4 enabled - a whooping 157 fps (+21% over 3.7 L4 disabled)
CPU 4.2 L4 enabled - an even more whooping 175 fps (+24% over 4.2 L4 disabled)

Well that's what an L4 is, technically I guess. What would be interesting to see is if the RAM speeds also have a big effect on any of the tasks, I mean how does OC L4 + stock RAM vs stock L4 + OC RAM results look like? I'm talking about 2100 MHz or above, in terms of RAM speeds.

Also how does the OCed 5775c handle OCing L4 & RAM, whether there's any benefit doing all 3 or just OCing any 2 of them is better?

 

[Nordic]

The reviews that I read that specifically looked at the edram say it provides marginal improvements.
http://www.legitreviews.com/edram-overclocking-on-the-intel-core-i7-5775c-broadwell-cpu_167425
http://www.pcworld.com/article/2950...ted-desktop-broadwell-has-one-neat-trick.html

 

[cucker tarlson]

The reviews that I read that specifically looked at the edram say it provides marginal improvements.
http://www.legitreviews.com/edram-overclocking-on-the-intel-core-i7-5775c-broadwell-cpu_167425
http://www.pcworld.com/article/2950...ted-desktop-broadwell-has-one-neat-trick.html

Well I guess they suck then. How can you miss a 30% improvement ? I've seen this problem back when I became interested in the topic of ram speed vs min. fps. Most reviewers are just too lazy or incompetent to use a proper methodology. They just run a built-in game benchmark, or worse (oh dear God) a GPU benchmark like valley, report average fps increase of 1% between slow and fast ram, and then we've got the community totally misled unless someone posts their own results or a smart reviewer takes time to do it properly and gets enough credit.

Well that's what an L4 is, technically I guess. What would be interesting to see is if the RAM speeds also have a big effect on any of the tasks, I mean how does OC L4 + stock RAM vs stock L4 + OC RAM results look like? I'm talking about 2100 MHz or above, in terms of RAM speeds.

Also how does the OCed 5775c handle OCing L4 & RAM, whether there's any benefit doing all 3 or just OCing any 2 of them is better?

will try soon But pretty sure all of them have a big impact.

frequency increase translates more or less proportionally to added performance, I measured 12% fps increase from 14% oc (3700->4200). Then on top of that additional 21% from using stock 1.6Ghz L4, make it 32% when L4 is oc'd to 2GHz. Will do a ram test in a moment.

Tried to do the test but my bios is acting up and I am getting fed up. Will attempt to test that at a later time.

 

[OneMoar]

thread ripper could benefit from edram they are already wasting a bunch of package space

 

[Aquinus]

It wasn't about it making the CPU too good, it was entirely because of cost. The eDRAM was a second die on the CPU package, and that die was about 50% of the size of the CPU die. Die space is what makes a CPU expensive to manufacture, and multiple dies on a package adds even more expense.

It's also eDRAM, not SRAM. DRAM is far cheaper to manufacture than SRAM like standard cache is because all you're talking about is a transistor and a capacitor versus 6 transistors for a single bit. It's easier to produce DRAM because there is literally less circuitry to build it. Controlling DRAM is the tricky part but, it's not like Intel doesn't already know how to do that with DRAM that isn't on the same package.

Simply put, putting DRAM on the package is smart because you get the benefit of DRAM which is cost and the benefit of cache which is distance from the "processor" (in the sense of the thing doing the processing, like a core or a hardware thread.) The extra cost would probably come from wiring up everything on the package itself, not the production of the eDRAM.

thread ripper could benefit from edram they are already wasting a bunch of package space

This, this, and more this.

 

[Vario]

If I recall correctly, The 5775C didn't clock as high as the 4790K, Sky Lake was right around the corner, and the 5775C release wasn't really promoted for gaming. There were some Z97 boards that didn't support it and it didn't work with Z87. However, there were plenty of reviews in 2015 extolling the virtues of the Crystal Well EDRAM for gaming.

The best time to buy it was in 2015-2016 when it was in the upper $200s. Now as it is in the upper 300s low 400s, it is priced so high its not worth buying it over a Kabylake which has it on typical overclocked clock speed and lets you run a modern chipset, ddr4, etc.

I see the i5 5675C is in the mid-upper $200 range so it might be alright if you already have a 4690K or a 4790K, a Z97 board that is compatible, and don't mind losing 2 threads if 4790.

 

[biffzinker]

The best time to buy it was in 2015-2016 when it was in the upper $200s. Now as it is in the upper 300s low 400s, it is priced so high its not worth buying it over a Kabylake which has it on typical overclocked clock speed and lets you run a modern chipset, ddr4, etc.

Had a look at the historical pricing for the 5775C, not seeing anything were it dipped under $300.
PC Part Picker

Also the end of life status shows the last shipments were in August.
End-of-Life date: Last order date is February 24, 2017
Last shipment date: August 11, 2017

Edited

 

[LAN_deRf_HA]

5775C didn't clock as high as the 4790K

That was part of the appeal, it was topping gaming charts while running much slower/using less power.

 

[Vario]

Had a look at the historical pricing for the 5775C, not seeing anything were it dipped under $300.
PC Part Picker
View attachment 91530

Also the end of life status shows the last shipments were in August.
End-of-Life date: Last order date is February 24, 2017
Last shipment date: August 11, 2017

Edited

Seemed like it was like $280 at microcenter IIRC.

 

[EarthDog]

Did i miss wjere response in one title makes the rest true? I recall reviews on these and didnt see it making much of a difference in games at the time.. it didnt in our testing.

What changed...and what other games are we seeing positive results in?

 

[LAN_deRf_HA]

Did i miss wjere response in one title makes the rest true? I recall reviews on these and didnt see it making much of a difference in games at the time.. it didnt in our testing.

What changed...and what other games are we seeing positive results in?

As long as I saw it in reviews it was always at the top or near it against much faster running chips, even in non cpu limited scenarios.

And on a side note it also hasn't been topped on integrated graphics either.

 

[EarthDog]

Something like that shows parity across many chips. Looking at the review where it came from, its up there, but, you arent seeing differences like what the op is seeing is my point. That isnt common.

http://www.anandtech.com/show/9320/intel-broadwell-review-i7-5775c-i5-5675c/9

Thread title should add the game for clarity so as not to infer this happens in all, or even many, games.

 

[eidairaman1]

Well this does nothing more than prove if you have a sandy/ivy/has/broad, there is no point in upgrading to sky/kaby...

 

[Nordic]

How can you miss a 30% improvement ?

I am not saying that there is no performance improvements. Legit reviews showed a 1% improvement which does sound too small. DDR3 memory scaling on haswell had as high as 12% more performance with faster memory in certain games. This edram should at least show 12% in certain memory intensive games, but 30% across the board sounds outlandish. If edram increased performance 30% across the board, I would think intel would have more processors with it despite the cost.
http://www.anandtech.com/show/7364/memory-scaling-on-haswell/7

 

[DeathtoGnomes]

@cucker tarlson ok this is all good info, but what what about power consumption, whats the difference with your tweaking here?

 

[cucker tarlson]

Something like that shows parity across many chips. Looking at the review where it came from, its up there, but, you arent seeing differences like what the op is seeing is my point. That isnt common.

http://www.anandtech.com/show/9320/intel-broadwell-review-i7-5775c-i5-5675c/9

Thread title should add the game for clarity so as not to infer this happens in all, or even many, games.

tested on 290X/980 and you're surprised the games aren't showing differences? There's a gpu bottleneck in all of them,lol,can't you see that ?
Title is as clear as day, it says what I found about the performance. How can that be clearer. Mention the games ? Uhm, all of them ? Did you read the review from purepc where they test on 1080p GTX 1080 in cpu limited parts of the game ? Try that maybe. Here's a teaser

My second post of the thread (page 1, "spis treści" is the table of contents, click to unfold). What's funny is the the reviewer,who always engages in discussion below his articles, said in the oc vs oc test he overclocked the 5775c core only, edram was left at stock. With what I assume is ~10% difference between stock (1.6) and 2GHz the 5775c should sail way clear of 7700K.

@cucker tarlson ok this is all good info, but what what about power consumption, whats the difference with your tweaking here?

L4 has no impact on power draw. Core oc does but the cpu sips power anyway. The package wattage reading at 4.2GHz 1.29v with L4 2GHz OC is still around 60W max. IBT very high raises it to around 80W.

 

[biffzinker]

The Euler3d CFD Benchmark shows what the 5775C is capable of compared to a overclocked 6700K (4.6 GHz.)

http://www.legitreviews.com/intel-core-i7-6700k-skylake-processor-review_169935/7

There is also overclocking the eDRAM.

http://www.legitreviews.com/edram-overclocking-on-the-intel-core-i7-5775c-broadwell-cpu_167425

 

[cucker tarlson]

I am not saying that there is no performance improvements. Legit reviews showed a 1% improvement which does sound too small. DDR3 memory scaling on haswell had as high as 12% more performance with faster memory in certain games. This edram should at least show 12% in certain memory intensive games, but 30% across the board sounds outlandish. If edram increased performance 30% across the board, I would think intel would have more processors with it despite the cost.
http://www.anandtech.com/show/7364/memory-scaling-on-haswell/7

I misrepresented even my own findings in this post, sorry. 30% in only for oc'd EDRAM (2GHz vs stock 1.6). Stock is around 20-25% when the cpu is the limitation but the gpu has a lot of horsepower left.

 

[EarthDog]

tested on 290X/980 and you're surprised the games aren't showing differences? There's a gpu bottleneck in all of them,lol,can't you see that ?
Title is as clear as day, it says what I found about the performance. How can that be clearer. Mention the games ? Uhm, all of them ? Did you read the review from purepc where they test on 1080p GTX 1080 in cpu limited parts of the game ? Try that maybe. Here's a teaser

My second post of the thread (page 1, "spis treści" is the table of contents, click to unfold). What's funny is the the reviewer,who always engages in discussion below his articles, said in the oc vs oc test he overclocked the 5775c core only, edram was left at stock. With what I assume is ~10% difference between stock (1.6) and 2GHz the 5775c should sail way clear of 7700k.

I see blanket assumptions applied. Its not a huge leap, but just knkwing some titles dont scale with intel clockspeeds, may curb some of the increases.

 

[wietjunk]

Hey i am new here, thanks for the awsome thread i have got this cpu for som time in the room sucking dust.

There is almoste nothing about this cpu on the net, tomorrow i receive my Asus rog impact VII back from rma.

 

[eidairaman1]

There is almoste nothing about this cpu on the net, tomorrow i receive my Asus rog impact VII back from rma.

Get the latest release or beta firmware for it.

By the way it is against forum rules to post several replies with no one else's in between, so use the edit button and delete buttons accordingly. There is a multiquote button as well

 

[wietjunk]

Ok thanks.

 

[RichF]

Intel’s Skylake lineup is robbing us of the performance king we deserve. The one Skylake processor I want is the one that Intel isn't selling.

 

[eidairaman1]

Ok thanks.

Combine your posts above please